Proportional electoral formulae minimize the difference between the vote and seat share of each party but some methods are considered `more proportional' than others. The reason can usually be attributed to the vagueness on what should be considered as a correct measure of proportionality. In fact, electoral formulae and measures of disproportionality are strictly related: each formula yields a seat assignment which minimizes a precise disproportionality index. Thus, the common procedure of testing proportionality of all methods using one of such indexes can be misleading. We propose to evaluate the proportionality of electoral formulae in terms of their robustness. The results show that while no method dominates the others in terms of proportionality, Sainte-Laguë and LR-Hare tend to be `robust' respect to the set of indexes here chosen.