Historically, it has often been asserted that most stars form in compact clusters. In this scenario, present-day gravitationally-unbound OB associations are the result of the expansion of initially gravitationally-bound star clusters. However, this paradigm is inconsistent with recent results, both theoretical and observational, that instead favour a hierarchical picture of star formation in which stars are formed across a continuous distribution of gas densities and most OB associations never were bound clusters. Instead they are formed in-situ as the low-density side of this distribution, rather than as the remnants of expanding clusters. We utilise the second $Gaia$ data release to quantify the degree to which OB associations are undergoing expansion and, therefore, whether OB associations are the product of expanding clusters, or whether they were born in-situ, as the large-scale, globally-unbound associations that we see today. We find that the observed kinematic properties of associations are consistent with highly substructured velocity fields and additionally require some degree of localised expansion from sub-clusters within the association. While most present-day OB associations do exhibit low levels of expansion, there is no significant correlation between radial velocity and radius. Therefore, the large-scale structure of associations is not set by the expansion of clusters, rather it is a relic of the molecular gas cloud from which the association was formed. This finding is inconsistent with a monolithic model of association formation and instead favours a hierarchical model, in which OB associations form in-situ, following the fractal structure of the gas from which they form.