christopher hennessy | professor of finance | london business school

prof.christopher.hennessy@gmail.com

chennessy@london.edu

my research interest is in assessing the meaning and content of empirical tests. a good way to assess "credibility" is to create a model laboratory and mimic the tests performed by empiricists in seemingly-ideal settings free of the types of noise found in econometrics texts and argued about obsessively in seminars. what are we actually learning from the accumulation of all the evidence-- even that achieving randomization? my goal is to clearly understand challenges and limitations by way of parable economies, with agents sometimes being employed as econometricians. hopefully, you will find the papers stimulating even if we disagree! and hopefully some of this agenda will be taken up by others. things are moving forward, but we must be self-critical if we are to progress. for example, "equilibrium counterfactuals" is a critique of structural moment matching methods, an approach used in some of my papers. still lots to learn and do!

newer papers...

signaling, instrumentation, and CFO decision-making, with Gilles Chemla, JFE

Building parable economies embedding econometricians, we view alternative estimators (IV, fuzzy RD, natural experiments, OLS, event studies) from the perspective of privately-informed decision-makers, e.g. CFOs. IV estimates can be misleading since randomization through observable instruments eliminates signal content arising from discretion. If the goal is informing discretionary decisions, rather than predicting outcomes after forced/mistaken actions, instrumentation is problematic, whereas OLS or event studies can be sufficient. The analysis shows the utility of alternative estimators hinges upon oft-neglected assumptions about agent/econometrician information sets, as distinct from exclusion restrictions. We recommend parable economy estimation as precursor to real-world IV estimation.

markets versus mechanisms, with ralph boleslavsky and david kelly, RFS

We demonstrate limitations on usage of direct revelation mechanisms (DRMs) by corporations inhabiting economies with securities markets. We consider a corporation seeking to acquire decision relevant information. In an environment with a securities market, the act of posting a standard DRM endogenously increases the informed agent's outside option value. If the informed agent rejects said DRM, he convinces the market he is uninformed, and he can trade aggressively with low price impact, generating large off-equilibrium trading gains, which increases his outside option. Due to this endogenous outside option value effect, using a DRM to screen out uninformed agents may be impossible. Moreover, even when screening is possible, refraining from posting a mechanism and instead relying on markets for information is optimal if the endogenous change in outside option value is sufficiently large. Finally, even when posting a DRM dominates relying on markets, superior outcomes are achieved by introducing a search friction which randomly limits the agent's ability to observe the DRM, forcing the firm to sometimes rely on markets for information.

goodhart's law and machine learning: a structural perspective, joint with Charles A.E. Goodhart, accepted, IER

We develop a structural framework illustrating how penalized regression algorithms affect Goodhart bias when training data is clean but covariates are manipulated at known cost by future agents facing prediction models. With quadratic manipulation costs, bias is proportional to the Ridge penalty term. In the limit under increasingly steep cost functions, bias is proportional to the Lasso penalty term. If costs depend upon either absolute or percentage manipulation, the following algorithm yields manipulation-proof prediction: Within training data, evaluate candidate coefficient vectors at their respective incentive-compatible manipulation configuration. We obtain analytical expressions for coefficient adjustments: slopes (intercept) shift downward if costs depend upon percentage (absolute) manipulation. Statisticians ignoring agent-borne manipulation costs select socially suboptimal penalization, resulting in socially excessive, and futile, manipulation. Model averaging, especially over Lasso or ensemble estimators, reduces manipulation costs significantly.

papers that are a bit older...

theory-based empirical testing and theory of empirical testing

equilibrium counterfactuals, forthcoming, international economic review (joint with gilles chemla)

We incorporate structural modellers into the economy they model. Using the traditional moment-matching method, they ignore policy feedback and estimate parameters using a structural model that treats policy changes as zero probability (or exogenous) "counterfactuals." Estimation bias occurs since the economy's actual agents, in contrast to model agents, understand policy changes are positive probability endogenous events guided by the modellers. We characterize equilibrium bias. Depending on technologies, downward, upward, or sign bias occurs. Potential bias magnitudes are illustrated by calibrating the Leland (1994) model to the Tax Cuts and Jobs Act of 2017. Regarding parameter identification, we show the traditional structural identifying assumption, constant moment partial derivative sign, is incorrect for economies with endogenous policy optimization: The correct identifying assumption is constant moment total derivative sign accounting for estimation-policy feedback. Under this assumption, model agent expectations can be updated iteratively until the modellers' policy advice converges to agent expectations, with bias vanishing.

learning, parameter drift, and the credibility revolution, journal of monetary economics, 2020 (joint with dima livdan)

This paper analyses interpretation and extrapolation of evidence from natural experiments in dynamic economies when underlying data generating processes are latent. Endogenous belief updating results in parameter drift: Shock response signs and magnitudes vary widely over time despite all shocks being ideally exogenous. Moreover, learning causes otherwise symmetric shock responses to become asymmetric, further limiting extrapolation. Closed-form formulae are derived for comparing shock responses at different points in time or inferring causal parameters (comparative statics and deep parameters). Correct interpretation and extrapolation hinges upon the time pattern of realized shocks and parametric assumptions regarding potential arrival intensities. It is proven that a martingale condition is necessary and sufficient for shock responses to directly recover comparative statics but stochastic monotonicity is insufficient to ensure correct recovery of signs.

controls, belief updating, and bias in medical rcts, journal of economic theory, 2020 (joint with gilles chemla)

Double-blind RCTs are viewed as the gold standard in eliminating placebo effects and identifying non-placebo physiological effects. We develop a formal model of placebo effects. If subjects in seemingly-ideal single-stage RCTs update beliefs about breakthroughs based upon personal physiological responses, mental effects differ across medications received, treatment versus control. Consequently, the average cross-arm health difference becomes a biased estimator. Constructively, we show: bias can be altered through choice of control; higher-efficacy controls mitigate upward bias; and efficacy states can be revealed through controls of intermediate efficacy or controls that mimic a subset of efficacy states. Consistent with experimental evidence, our theory implies outcomes within-arm and cross-arm differences can be non-monotone in treatment probability. Finally, we develop a novel differences-in-differences test to detect RCT bias.

beyond random assignment: credible inference and extrapolation in dynamic economies, journal of finance, 2020 (joint with ilya strebulaev)

even if randomization is granted, how severe are biases due to dynamic policy uncertainty? we offer analytically tractable formulae for bias signs/probabilities. we also derive a set of identifying assumptions for correct causal effect signs and magnitudes. the assumption of unanticipated permanent shocks is neither necessary nor sufficient. a weaker condition, martingale policy, is necessary and sufficient.

rational expectations and the paradox of policy relevant natural experiments, journal of monetary economics, 2019 (joint with gilles chemla)

evidence from randomization is contaminated by ex post endogeneity if it is used to set policy endogenously in the future. measured effects depend on objective functions into which experimental evidence is fed and prior beliefs over the distribution of parameters to be estimated. endowed heterogeneous effects generates endogenous belief heterogeneity making it difficult/impossible to recover causal effects. observer effects arise even if agents are measure zero, having no incentive to change behavior to influence outcomes.

empirical analysis of corporate tax reforms: what is the null and where did it come from?, journal of financial economics, 2019 (joint with akitada kasahara and ilya strebulaev)

empiricists analyzing corporate responses to tax reforms have been forced to rely on comparative statics from constant tax rate models--which do not speak to the data generating processes being exploited. such comparative statics exercises predict large symmetric responses to tax rate changes. to fill the theoretical void we solve analytically a dynamic model of optimal leverage in which the tax rate follows a two-state markov chain. in this setting, corporations are more responsive to tax rate increases than to decreases. in simulations of ideal diff-in-diff estimation, the known-to-be-false null that taxes do not matter, cannot be rejected roughly half the time. further, regression coefficients decline even as tax-induced deadweight losses rise as we increase bankruptcy cost parameters. this is another paper illustrating the probable problem of false-falsification and the non-obvious interpretation of coefficients. it is premature to reach consensus regarding tax effects.

tobin's q, debt overhang and corporate investment, journal of finance, 2004 [Brattle Prize]

how to take myers (1977) to the data. what regressor does the theory imply?

debt dynamics, journal of finance, 2005 [Brattle Prize], with Toni Whited

illustrates the false-falsification of trade-off theory by running mimicking regressions using a simulated trade-off theoretic firm

how costly is external financing: evidence from a structural estimation, journal of finance, 2007 [Brattle Prize], with Toni Whited

simulated moment estimation. investment-cash flow sensitivity non-mono in financing costs. economic meaning of constraint proxies.

testing q theory with financing frictions, 2007, journal of financial economics, with toni whited

extending hayashi (ecma, 1982) to incorporate mm violations. how far can we get analytically? functional form assumptions needed?

repeated signaling and firm dynamics, 2010, review of financial studies, with Dima Livdan

how to take a least-costly separating equilibrium (one equilibrium in myers-majluf) to the data. endogenous financing and investment.

model before measurement, 2012, critical finance review (invited essay)

attempts to clarify what the hennessy-whited papers said and meant: we don't know the true theory, if there is one, but verbally plausible arguments are no way to assess what theory predicts about the dynamic behavior of leverage ratios and investment rates.

theories of optimal capital structure and security design

can the tradeoff theory explain debt structure?, review of financial studies, 2007, with dirk hackbarth and hayne leland

optimal debt structure places bank senior to public debt to max flexible debt capacity with junior public debt featuring standard tradeoff.

why does financial structure vary with macroeconomic conditions?, journal of monetary economics, 2007, with amnon levy

ge model with agency problems creating ceiling on leverage and floor on equity stake so internal equity drives dynamics.

taxation, agency conflicts, and the choice between callable and convertible debt, journal of economic theory, 2008, with yuri tserlukevich

stochastic differential game with instantaneous volatility choice and infinite sequence of levered recaps. "modern" extension of rick green's early paper on the topic (1984)

debt, bargaining, and credibility in firm-supplier relationships, journal of financial economics, 2009, with dima livdan

debt is bargaining tool depleting bilateral firm-supplier bilateral surplus ex ante, but this compresses the set of credible implicit contracts ex post.

acquisition values and optimal financial inflexibility, journal of financial economics, 2010, with uli hege

there is a cost for incumbent in keeping deep pockets, as it encourages entry for buyout. empty pockets my deter entry but also leaves you defenseless if entry occurs.

a theory of debt market illiquidity and leverage cyclicality, review of financial studies, 2011, with josef zechner

kyle-type debt trading with endogenous uninformed trading. a liquid debt market enables large debt stakes, facilitating ex post renegotiation. multiple equilibria.

skin in the game and moral hazard, journal of finance, 2014, with gilles chemla

bank hidden effort ex ante and hidden information ex post. optimal retentions feature common min jr stake (pooling) or menu of jr stakes (separating).

secondary market liquidity and security design: theory and evidence from abs markets, review of financial studies, 2016

kyle meets security design (tractable). what way of packaging cash flow minimizes expected trading losses and carrying costs if the uninformed hit with liquidity shocks and trade endogenously? the theory predicts multiple tranches with senior being more liquid. in contrast to claims on corporate cash flows, in abs markets, more senior claims are in fact more liquid, consistent with the theory.

learning and leverage dynamics in general equilibrium, 2016, review of finance, with boris radnaev

ge setting with tradeoff theoretic firms optimizing in economy with unobserved "disaster" risk. with rational bayesian learning we should expect leverage run-up during great moderations, and sharp reactions to disaster realizations. these are not a priori evidence in favor of government interventions or irrationality.

government as borrower of first resort, 2016, journal of monetary economics, with gilles chemla

government debt siphons off uninformed demand for corporate debt, reducing endogenous gains to informed trading, lowering info quality and potentially pruning pooling at high debt or increasing welfare in the event of such pooling.