This paper presents a method to predict the uncertainty associated with dynamic identification of a reduced-scale model of a building structure from laboratory experiments. Systematic repetitions of dynamic identification experiments and their results were expanded through Hermite Polynomial chaos. The study describes the degree of uncertainty related to structural engineering properties of the model due to laboratory errors. Uncertainties in the physical experimental system should be adequately modelled to obtain an accurate estimate of structural safety. This is crucial to reduce prediction errors of full-scale structural performance. The method works using reliability-based optimization. Applicability of the framework is illustrated using a reduced-scale model of a high-rise building, constructed for multi-hazard dynamic experiments on a shaking table and in a wind tunnel. The still-air dynamic identification of model parameters closely affects the wind tunnel results. The method accounts for system variability and experimental error propagation, enabling evaluation of structural reliability for the corresponding full-scale structure. Results reveal that the variability introduced by laboratory errors propagates onto the estimation of structural frequencies and damping, leading to non-negligible variation of full-scale structural parameters. The corresponding probability density function of some relevant-mode full-scale vibration magnitudes is discussed as a compelling example for structural safety. The study demonstrates that experimental variability must be taken into account, particularly in the case of aeroelastic wind tunnel models, because structural response measurements are affected by this uncertainty.