The authors investigate the evolution of debt policies of business startups over a period of 15 years. Based on a sample of Belgian nonfinancial startups founded between 1996 and 1998, they find that the debt policies of these firms are extremely stable over time and that the initial debt policy is an important predictor of the future debt structure. The founding CEO has a large impact on the stability of debt policies: When a new CEO takes over, the link between initial and future debt policies is significantly weakened.

An important aspect in the valuation of listed and private companies is the current and expected debt policy. There are competing theories about the development of debt policies. Information-based theories predict that these policies will change as firms mature, whereas imprinting theory suggests that founding decisions play an important role in the determination of firm characteristics, which are perpetuated over time.

The unique dataset of close to 5,000 Belgian startups allows the authors to test these competing theories. Their empirical evidence is consistent with imprinting theory; the initial debt policy is important in explaining future debt policies. This link is significantly weaker in cases when the founding CEO departs or passes away. The authors thus consider the preferences of the founding CEO to be a highly relevant, firm-specific factor that helps to explain consistency in the debt structure over time.

The authors collect data on 4,962 Belgian limited liability firms that were incorporated in the years 1996–1998. They follow these firms until 2013. About half of the firms are active in all sample years, whereas the other half of the firms leave the sample because of bankruptcy, acquisitions, or buyouts. In addition, the authors collect data about the departure or death of the founding CEO for all firms from the Belgian Law Gazette.

Four distinct debt measures are constructed: (1) the leverage ratio, (2) debt specialization (i.e., the choice among five different types of debt financing), (3) debt maturity, and (4) debt granularity (i.e., the extent to which a firm spreads out the maturity dates of its debt). To evaluate the role of the initial founder, two dummy variables capturing CEO departure or death are constructed. Control variables include firm size, profitability, tangibility, growth opportunities, capital expenditures, and creditworthiness. Where appropriate, year and industry fixed effects are included.

To determine the influence of initial conditions on future debt policies, the authors estimate increasingly complex sets of regressions for all four measures of debt policy. The results show the existence of an important stable component. Next, the authors use variance decompositions to consider the importance of time-invariant and firm-specific effects. They document the relevance of an unobserved firm-specific factor that drives debt policies. Finally, the authors use a set of dummy variables to test for the influence of founding CEO departures on the evolution of debt policies. They conclude that founding CEOs imprint their mark on the debt policies of the startups.

The authors also evaluate a number of alternative explanations for the empirical regularities discovered, which can be ruled out based on additional econometric testing. These alternative rationalizations include (1) slow speeds of adjustment toward the target debt policy, (2) the mere presence—as opposed to preferences—of the founding CEO, (3) stability of the debt structure as a function of information asymmetry, (4) heavy influence of slow-growing firms on the results, (5) heavy influence of small firms on the results, and (6) biases introduced by firms that exit the sample.

The authors enrich the capital structure debate by providing novel time-series evidence consistent with imprinting theory. Even though it is probably no easy matter to create datasets of similar breadth and quality for other countries, such an attempt would be valuable in order to provide a fuller basis for confirming or challenging the authors’ findings.