The Federal Reserve's Corporate Credit Facilities (CCFs) were launched in early 2020 amid significant volatility in the U.S. corporate bond market. The CCFs promised both direct support to firms via cash bond purchases, as well as indirect support via purchases of exchange-traded funds (ETFs). In this paper, we provide estimates of the treatment effect on corporate bond spreads from direct cash bond support by the CCFs. To do so, we introduce a novel identification strategy that exploits the ratings heterogeneity of corporate bonds across firms. We estimate that the initial announcement on March 23, 2020 of the CCFs led to a 96 bps decline in eligible issuers' spreads. To estimate the effect of the announced expansion of the facilities on April 9, 2020, we exploit a quasi-natural experiment. Fallen Angel issuers were initially eligible for the CCFs and fell out of eligibility but then had their eligibility reinstated at the same time as the expansion announcement. We compare these issuers with a comparable control group and find that the treatment effect for the expanded size of the facilities is -126 bps. Using a novel causal machine learning approach, we estimate the counterfactual treatment effect for ineligible issuers had they received direct cash bond support (and additional indirect support via ETFs) on March 23, 2020 to be around -500 bps. While large, this estimate appears plausible considering that the spreads of Fallen Angel issuers tightened around 300 bps on April 9, 2020 when their eligibility was restored.
Heterogeneous Treatment Effects and Counterfactual Policy Targeting Using Deep Neural Networks
An Application to Central Bank Corporate Credit Facilities
2025
Develops a semi-parametric DiD estimator using deep neural networks for dynamic heterogeneous treatment effects.
Demonstrates: Methodological innovation combining machine learning with causal inference, applied to policy-relevant questions.
Key Contributions
Presents a novel two-step semi-parametric difference-in-differences estimator
Uses deep neural networks for non-parametric terms with high-dimensional controls
Shows CCFs increased leverage and payouts but failed to boost investment
Methods
Deep neural networks for propensity/outcome modeling
Semi-parametric difference-in-differences
Dynamic heterogeneous treatment effect estimation
I present a novel two-step semi-parametric difference-in-differences (DiD) estimator for computing dynamic (heterogeneous) treatment effects and policy counterfactuals. In the first step, deep neural networks are used to compute non-parametric terms in a setting with high-dimensional controls. These are inputs into the estimator evaluated in the second step. The estimator is applied to study the effects of the Federal Reserve's Corporate Credit Facilities (CCFs) on the dynamics of firm cash holdings, leverage, payout, and investment. I show that the proposed estimator delivers comparable results to static (homogeneous) treatment effects obtained from DiD panel regressions and dynamic (homogeneous) treatment effects from event study regressions with two-way fixed effects, though with important differences attributable to selection bias and heterogeneity. Firms generally increased cash holdings and leverage, while payout and investment initially fell. Firms eligible for the CCFs accumulated less cash and began deleveraging in 2021, relative to ineligible firms. Eligible firms exhibit relatively larger payouts, while they do not invest more, suggesting that the CCFs failed to meet their objective of boosting real effects. Counterfactual treatment effects provide mixed to inconclusive evidence that expanding eligibility of direct cash bond support from the CCFs would have improved investment outcomes while providing stronger evidence that firms would have increased leverage and payouts.
Central Bank Corporate Bond Purchase Programs: Commitment Matters
2025
Shows why secured vs. unsecured debt intervention leads to different firm outcomes using dynamic capital structure theory.
Demonstrates: Rigorous theoretical modeling with clear policy implications.
Key Contributions
Novel observation that ECB and Fed engaged in de facto unsecured debt intervention
Explains why targeted firms increased leverage/payouts but not investment
Demonstrates secured debt intervention could produce better outcomes via commitment
Methods
Dynamic capital structure model with investment
Theoretical analysis of commitment mechanisms
Comparison of ECB and Federal Reserve programs
Over the past decade, the European Central Bank (ECB) and the Federal Reserve expanded the limits of unconventional monetary policy to directly provide firms with financing through corporate bond purchases. Empirical research has found that these programs led to increased leverage for directly targeted firms, as well as relatively higher payouts to shareholders but no relative increase in investment, contrary to the central banks' stated objectives. This paper makes the novel observation that both the ECB and Fed engaged in de facto unsecured debt intervention in financially unconstrained firms. I show that the stated stylized empirical facts arise in a dynamic capital structure model with investment where firms lack commitment to an ex ante debt policy. Unsecured debt intervention accelerates debt issuance to such an extent that higher potential debt prices are completely offset by increased leverage. Moreover, rather than being used for investment, the proceeds are distributed to shareholders. In contrast, secured debt intervention results in more favorable credit and investment dynamics, even among financially unconstrained firms. Secured debt issuance is disciplined by the collateral constraint, which induces commitment, thus allowing firms to benefit from intervention.
The Effects of Design of Central Bank Corporate Credit Facilities
PhD Dissertation, University of Chicago Booth School of Business, 2025
The three papers above comprise this dissertation, examining the effects, heterogeneity, and theoretical mechanisms
of central bank corporate bond purchase programs.
Demonstrates: Sustained research program integrating empirical and theoretical methods.