Until now, it has been impossible to observationally measure how star cluster scaleheight evolves beyond 1Gyr as only small samples have been available. Here, we establish a novel method to determine the scaleheight of a cluster sample using modelled distributions and Kolmogorov-Smirnov tests. This allows us to determine the scaleheight with a 25% accuracy for samples of 38 clusters or more. We apply our method to investigate the temporal evolution of cluster scaleheight, using homogeneously selected sub-samples of Kharchenko et al. (MWSC, 2012, Cat. J/A+A/543/A156, 2013, J/A+A/558/A53 ), Dias et al. (DAML02, 2002A&A...389..871D, Cat. B/ocl), WEBDA, and Froebrich et al. (FSR, 2007MNRAS.374..399F, Cat. J/MNRAS/374/399). We identify a linear relationship between scaleheight and log(age/yr) of clusters, considerably different from field stars. The scaleheight increases from about 40pc at 1Myr to 75pc at 1Gyr, most likely due to internal evolution and external scattering events. After 1Gyr, there is a marked change of the behaviour, with the scaleheight linearly increasing with log(age/yr) to about 550pc at 3.5Gyr. The most likely interpretation is that the surviving clusters are only observable because they have been scattered away from the mid-plane in their past. A detailed understanding of this observational evidence can only be achieved with numerical simulations of the evolution of cluster samples in the Galactic disc. Furthermore, we find a weak trend of an age-independent increase in scaleheight with Galactocentric distance. There are no significant temporal or spatial variations of the cluster distribution zero-point. We determine the Sun's vertical displacement from the Galactic plane as Z_{sun}_=18.5+/-1.2pc.