Abstract The relationship between production in total manufacturing measured as an ordinary volume series and various barometer series from the Swedish Business Tendency Surveys is investigated using data from the period 1968–;1992. Models are constructed using a systematic strategy aiming at parsimonious models within the Autoregressive-Distributed Lag (ADL) class. Autoprojective models that serve as a baseline in the comparisons are estimated. The performance of these models is compared with models that include barometer series. In the main, seasonally unadjusted series are used and models formulated both in terms of quarter-to-quarter and annual changes are considered. As regards immediate ease of interpretation, the latter have an advantage, but there is not much difference between the quality of the two types of models. The best barometer series is found to be volume of production and a model including this series is a significant improvement on the best autoprojective model. There is a close relationship between the barometer production series and the annual change in the volume series. About twenty-five other barometer series are also considered, but they are not found to provide additional information. In order to come closer to a forecasting model, ex ante variables are also considered. The connection between each ex ante and the corresponding ex post variable is investigated including a study of possible lagged and seasonal effects. Finally models based on the information set available at time-point “ t -1” are obtained.