The 'trade-off theory' of capital structure predicts a positive relationship between earnings and leverage, contradictory to well established empirical evidence. Since corporate earnings are known to be mean reverting, we reformulate the trade-off model with mean reverting earnings. We show that, with mean reverting earnings, there should be a negative relationship between optimal leverage and the current earnings level. The model also illustrates the importance of the earnings reversion parameter in the determination of capital structure. The major implications of the model are supported by empirical tests carried out with a sample of firms in the S&P 500 Index.The capital structure decision is arguably one of the most important issues in corporate finance. Beginning with the irrelevance propositions of Miller and Modigliani (1958), there is a substantial literature on what Myers (1989, p. 80) calls 'the search for optimal capital structure'. In the theoretical literature, there are two major competing models, the traditional 'trade-off theory' and the more recent 'pecking order hypothesis'. According to the former, firms choose an optimal leverage ratio by balancing the tax benefits of debt against the costs associated with debt, e.g., bankruptcy costs or financial distress costs. The pecking order hypothesis is based on informational asymmetry between insiders (managers or shareholders) and outsiders. According to this theory, firms tend to raise funds in the following order of preference: internal finance (retained earnings), safe debt, risky debt, and finally equity.Certain implications of the trade-off theory have received empirical support; see, for example, Long and Malitz (1985), Mackie-Mason (1990) and Graham (1996). However, in a head-to-head comparison between the two theories, the pecking order model seems to be more favourably viewed than the trade-off theory (Baskin, 1989;Shyam-Sunder and Myers, 1999;Fama and French, 1999). To quote Fama and French (1999, p. 3) '… we identify one big contradiction of the trade-off model and one less serious contradiction of the pecking order model.' The 'big contradiction' is the earnings-leverage relationship. The actual relationship between leverage and earnings (or profitability) has been found to be negative (Titman and Wessels, 1988;Baskin, 1989;Rajan and Zingales, 1995;Fama and French, 1999). But according to the received wisdom, the trade-off theory predicts a positive relationship between leverage and earnings; see Brealey and Myers (1996), Shyam-Sunder and Myers (1999) etc. To quote Myers (1989, 'the * This paper was presented at the 2001 Rutgers Conference on Capital Structure, Lancaster University, McMaster University, and University of Southern California. We appreciate comments received from participants at these seminars, as well as Peter McKay, Mark Weinstein and especially David Mauer. We also thank two anonymous referees for their feedback and for many useful suggestions, and the editor Mike Wickens for his guidance. Finally, we thank John Graham...