Final Report Summary - DYNACORP (Dynamic Structural Corporate Finance: Linking Theory and Empirical Testing)
The title of this European Research Council funded project is Dynamic Structural Corporate Finance: Linking Theory and Emprical Testing. The overarching objective of all projects subsumed in the grant it to use explicit formal models to better understand the meaning of empirical evidence in applied micro-econometrics, with a particular emphasis on corporate financial and investment decisions. A variety of research papers were produced shedding light on this issue from a variety of angles.
The paper, “Beyond Random Assignment: Credible Inference of Causal Effects in Dynamic Economies,” joint with Ilya Strebulaev of Stanford University, analyzes the meaning of econometric evidence derived from pure randomized experiments in dynamic settings. We show that random assignment is insufficient to recover the causal effects economists seek to estimate. We characterize the bias magnitudes and probabilities analytically. After showing that the biases are large, we derive a set of safe-harbor assumptions for correct identification of causal effect signs and magnitudes. Finally, we show how to recover causal effects from observed treatment responses in generic settings.
In “Corporate Responses to Exogenous Tax Changes” with Aki Kasahara and Ilya Strebulaev (both at Stanford University), we focus on randomized experiments featuring corporate taxes and leverage. This is an important policy item since tax authorities are very much concerned about the revenue and economic stability impacts of tax-induced increases in corporate leverage. The paper highlights to issues neglected in the empirical literature. First, we should that elasticities are not symmetrical, with firms increasing leverage more in response to tax increases than they decrease leverage in response to tax decreases. Second, we should that under plausible parameterizations, tax-induced responses to rate changes will appear to be statistically insignificant even in economies where tax concerns are the first-order determinant of leverage, illustrating the possibility of false-falsifications.
In “The Paradox of Policy-Relevant Natural Experiments,” joint with Gilles Chemla of Imperial College, we examine the interpretation of evidence drawn from a first-stage random assignment when that evidence will be used subsequently by the government in setting policy in the future. As we show, evidence from even ideal first-stage assignment becomes contaminated when it is used as an input into policy formation. After showing this problem, we show how the evidence can be corrected when such policy feedbacks are present.
In “Learning and Leverage Dynamics in General Equilibrium,” joint with Boris Radnaev of Cornerstone Research, we consider how firms should behave when they do not know the true nature of the random shocks they face. In particular, we consider firms that face an unobservable probability of disaster each period. These firms then rationally revise beliefs in a positive direction each period there is no disaster, but then revise beliefs downward when/if a disaster occurs. An important contribution of the paper is to show how the conjunction of learning, financing frictions and asset pricing dynamics are inter-linked. We argue the paper sheds light on behavior in the run-up to the financial crisis of 2007/8, as well as responses to the crisis.
In “Model Uncertainty and the Dynamics of Leverage and Investment,” joint with Andrea Gamba (Warwick) and Boris Radnaev (Cornerstone Research, formerly PhD student at LBS), we examine how firms should making their financing and investment decisions when they do not know the true nature of the shocks they face. In contrast to considering disaster risk, this paper considers that a firm faces an arbitrary number of different probability distributions and seeks to learn and optimize in such complex environments.
The questions addressed under the grant were central to an intense and stimulating research conference, Causal Inference in Finance and Economics, held at London Business School on 25 September 2015. At the conference participants debated best-practice in empirical testing and epistemology. Due to the strong positive response to the themes discussed, it is anticipated that this conference will become an annual event at LBS.
The paper, “Beyond Random Assignment: Credible Inference of Causal Effects in Dynamic Economies,” joint with Ilya Strebulaev of Stanford University, analyzes the meaning of econometric evidence derived from pure randomized experiments in dynamic settings. We show that random assignment is insufficient to recover the causal effects economists seek to estimate. We characterize the bias magnitudes and probabilities analytically. After showing that the biases are large, we derive a set of safe-harbor assumptions for correct identification of causal effect signs and magnitudes. Finally, we show how to recover causal effects from observed treatment responses in generic settings.
In “Corporate Responses to Exogenous Tax Changes” with Aki Kasahara and Ilya Strebulaev (both at Stanford University), we focus on randomized experiments featuring corporate taxes and leverage. This is an important policy item since tax authorities are very much concerned about the revenue and economic stability impacts of tax-induced increases in corporate leverage. The paper highlights to issues neglected in the empirical literature. First, we should that elasticities are not symmetrical, with firms increasing leverage more in response to tax increases than they decrease leverage in response to tax decreases. Second, we should that under plausible parameterizations, tax-induced responses to rate changes will appear to be statistically insignificant even in economies where tax concerns are the first-order determinant of leverage, illustrating the possibility of false-falsifications.
In “The Paradox of Policy-Relevant Natural Experiments,” joint with Gilles Chemla of Imperial College, we examine the interpretation of evidence drawn from a first-stage random assignment when that evidence will be used subsequently by the government in setting policy in the future. As we show, evidence from even ideal first-stage assignment becomes contaminated when it is used as an input into policy formation. After showing this problem, we show how the evidence can be corrected when such policy feedbacks are present.
In “Learning and Leverage Dynamics in General Equilibrium,” joint with Boris Radnaev of Cornerstone Research, we consider how firms should behave when they do not know the true nature of the random shocks they face. In particular, we consider firms that face an unobservable probability of disaster each period. These firms then rationally revise beliefs in a positive direction each period there is no disaster, but then revise beliefs downward when/if a disaster occurs. An important contribution of the paper is to show how the conjunction of learning, financing frictions and asset pricing dynamics are inter-linked. We argue the paper sheds light on behavior in the run-up to the financial crisis of 2007/8, as well as responses to the crisis.
In “Model Uncertainty and the Dynamics of Leverage and Investment,” joint with Andrea Gamba (Warwick) and Boris Radnaev (Cornerstone Research, formerly PhD student at LBS), we examine how firms should making their financing and investment decisions when they do not know the true nature of the shocks they face. In contrast to considering disaster risk, this paper considers that a firm faces an arbitrary number of different probability distributions and seeks to learn and optimize in such complex environments.
The questions addressed under the grant were central to an intense and stimulating research conference, Causal Inference in Finance and Economics, held at London Business School on 25 September 2015. At the conference participants debated best-practice in empirical testing and epistemology. Due to the strong positive response to the themes discussed, it is anticipated that this conference will become an annual event at LBS.