The Joint harmonized EU Investment survey, carried out for Italy by the Institute for Studies and Economic Analysis (ISAE), provides quantitative information about investment plans and the structure of investment and qualitative data on the factors influencing firm' behavior. As shown in the recent OECD' Business Tendency Surveys Handbook (OECD, 2003), in dealing with business surveys a special attention should be devoted to the control for the reliability of the results and to the appropriate use of weighting in processing survey' data. From a methodological standpoint, this is particularly important for surveys providing quantitative as well as purely qualitative data, as is the case for the Investment survey. In this respect, the paper proposes deterministic and statistically-based methods for the treatment of missing data and outliers, focussing especially on the quantitative question on investment plans. After a careful deterministic control, based on the possibility of re-interviewing non-responding firms, a fairly simple econometric model is used to estimate quot;residualquot; missing data, based on information on sector, geographical partition and number of employee of the firm, and its investment plans for the previous years (when available). The robustness of sample estimates in case of extreme observation in the tails of the distribution is then tested with the use of the winsorized mean. In processing the results, estimate of investment plans at industry level is then derived using sample weights, based, for each stratum, on the ratio between the Universe Population and that of the sample. Aggregation of qualitative data concerning investment structure and factors influencing firm' behaviour is then based on appropriate, variable-specific, size weights, i.e. the investment plan of the firm itself. This appears to be a fairly more appropriate choice with respect to the more usual value added-based weight. A presentation of the resulting data-bank, together with some consideration on possible future research, concludes the work
