In this paper we use the unit root test at both individual company (Dickey-Fuller) and panel (Im-Pesaran-Shin) level, in order to provide some quantitative evidence of the univariate behaviour of Italian companies' debt-ratio. If it is mean-reverting, then at least a share of companies are going to behave according to the trade-off model, whereas if it is non-stationary, then companies may behave according to pecking order theory. Individual company test results support the pecking order theory in the case of more than 80% of firms, while the panel test rejects the unit root null. These contradictory results can be explained by the heterogeneity of the firms which tends to characterise the entire panel. For this reason, we selected a number of sub-samples in which companies are supposed to follow either the pure trade-off or the pure pecking order model of behaviour. Results confirm that: a) heterogeneity may lead to a false rejection of the pecking order theory in panel unit root tests; b) both trade-off and pecking order theories contribute towards explaining the financial behaviour of Italian companies; c) the testing procedure we propose has the power to reject the pure pecking order model under the null hypothesis, but not the pure trade-off model.