The ISAE Market Services Survey: Methodological Upgrading, Survey Reliability, First Empirical Results

Within the E.U. Harmonised project, ISAE has been carrying out the quarterly services survey for Italy, since 1992, on a panel of about 1,000 business firms operating in the service sector. On E.U. suggestion, since the first quarter of 1998, the survey has been based on an updated sample, so as to enlarge the sectoral and information coverage. Starting from January 2003, the survey has been deeply revised and enlarged according to the Commission recommendations so as to cover the whole Market Service sector (Nace Rev.1.1 Sections H, I, J, K, part of O, with different degrees of detail). The survey is now based a sample of over 2,100 firms, stratified both by sector of activity (like already done since 1998) and geographic partitions; the units’ allocation to strata is optimal in the Neyman sense, with some integrations to deal with the pre-exiting panel and in order to include the most important Italian firms belonging to the target sectors. The aim of this paper is discussing the quality of the survey in terms of transparency and reliability. The methodological upgrading of the sampling design is hereafter discussed as well as each step of the new procedure. To begin with, the analysis of the Italian service sector is presented, which enables to set the necessary assumption in determining the survey frame, the selection of sampling units and the weighting system. The theoretical sampling design is thus explained with reference to sampling unit, stratification variables, units’ allocation to strata, as well as the various step of the whole procedure in terms of statistical treatment of non responses, weighting, aggregation steps sampling and non-sampling errors. Furthermore, using the first year of information collected, the remarkable gain in precision of the new sample is analysed, together with its reliability in terms of accuracy of estimates. The analyses of the design effect shows that, also when dealing with qualitative series, the Neyman allocation to strata considerably improves the precision of estimates in comparison both to the stratified proportional design and to the simple random one.