The authors demonstrate that corporate capital structures are not stable over time. Moreover, they show that capital structure stability is in fact the exception because it primarily occurs at low levels of leverage and it is almost always temporary. In addition, they find that industry median leverage levels, which are often used as target capital structure proxies, also vary over time. Target leverage models that place little or no weight on maintaining a particular leverage level are the best at modeling the instability of actual leverage cross-sections.

The majority of the academic literature supports the view that corporate capital structure is stable over time. The authors demonstrate a contradicting conclusion that has several implications for practitioners. They find substantial variation in leverage for publicly held industrial firms. If capital structure is not stable over time, then the current capital structure cannot reliably predict a future capital structure.

The authors also find substantial variation in industry median leverage ratios over time. Industry median leverage ratios are often used as target leverage proxies. But the authors admit that although these findings suggest that target leverage ratios change a lot over time, a closer look at the data indicates that there is a lot that is unknown about the time-series variation in leverage and the factors that generate such variation.

Finally, the authors show that extended periods of leverage stability arise on occasion but that permanently stable leverage is rare, with stable leverage typically occurring at low levels of leverage. In particular, they find that short-run stability goes away over horizons of 5–10 years and almost disappears over longer horizons. Moreover, they find that although it is observed reasonably frequently, high leverage is almost always temporary.

The authors analyze more than 15,000 industrial firms in the CRSP/Compustat database from 1950 to 2008. For their long-horizon analysis, they use two sets of data. The first is a subset of 2,751 firms with 20 or more years of data, and the second is a sample of 157 firms that have total assets data on Compustat from 1950 to at least 2000. In addition, the authors also analyze hand-collected leverage data that begin prior to the Great Depression for 24 firms in the DJIA.

As the measure of leverage, the authors consider book leverage, which they define as the ratio of total book debt to total book assets, and market leverage, which is defined as book debt divided by the sum of book debt plus the market value of common stocks. They ultimately select book leverage as their main measure. Their rationale is that both measures of leverage are highly correlated, which suggests little incremental information in the market leverage series, and that book leverage variation provides a lower bound on the instability in market leverage.

The authors then conduct a number of tests that revolve around various ways of splicing the data across different leverage levels and time. For example, one test determines the percentage of firms that had book leverage levels within a range not exceeding 5%, 10%, and 20% for a period of 10, 20, 30, and 40 years. Another test assesses the stability of the leverage cross-section. They test whether firms with high (or low) leverage in a given cross-section tend to have high (or low) leverage in future years’ cross-sections.

In addition, the authors perform simulations to determine which model among several candidate target leverage models explains the instability of leverage ratios over time. For each model, they average the many simulated iterations of the leverage cross-sections. They gauge the overall goodness-of-fit of each model by the sum of the root-mean-square error over a 20-year horizon plus the variation error, which penalizes models that generate cross-sectional instability because of greater leverage volatility than exists in the real data.

The idea that a firm’s capital structure or an industry’s median capital structure is stable over time has been the norm and pervades many real-world applications. For example, analysts often use a constant capital structure assumption based on the firm’s leverage or industry median leverage when calculating the weighted-average cost of capital. The results of the authors’ work make us think about the idea of an optimal capital structure. More importantly, what these results show and what the authors also allude to is that there is much that we do not know about the time-series variation in leverage.