Financial Heterogeneity and the Investment Channel of Monetary Policy

“Given the rich heterogeneity in financial positions across firms, a key question is: which firms are the most responsive to changes in monetary policy?” (Introduction, p. 2474). The paper answers that low-default-risk firms are the most responsive: “investment done by firms with low default risk is significantly and robustly more responsive to monetary policy than investment done by firms with high default risk” (p. 2474) – a finding the authors note is theoretically ambiguous ex ante, since financial frictions could either dampen high-risk firms’ responses (upward-sloping marginal cost of finance) or amplify them (via a financial-accelerator-style flattening of that cost curve).

The baseline specification interacts within-firm demeaned leverage or distance to default, x_{jt-1} - E_j[x_{jt}], with the monetary shock, rather than the raw level x_{jt}, “to ensure that our results are not driven by permanent heterogeneity in responsiveness across firms” (Section 2.2, eq. 2). This choice is motivated directly by the model, in which firms are ex ante homogeneous: “by demeaning financial position within firms… our estimates are instead driven by how a given firm responds to monetary policy when it has higher or lower default risk than usual” (Section 2.2). Firm and sector-by-quarter fixed effects, plus controls including the interaction of financial position with lagged GDP growth, further isolate the effect from other confounds.

“A firm has approximately a 0.7 unit lower semielasticity of investment to monetary policy when it is one standard deviation more indebted than it typically is,” and “approximately a 1.1 unit higher semielasticity when it is one standard deviation further from default than usual”; conditioning on both, “leverage is rendered statistically insignificant conditional on distance to default” (Section 2.2, Table III). The introduction states these differences “persist for up to 3 years after the shock and imply large differences in accumulated capital over time” (p. 2474).

“Consistent with the idea that default risk drives these heterogeneous responses, borrowing costs and the use of external finance increase by less for high-risk firms than for low-risk firms following a monetary expansion” (Introduction, p. 2474). Section 6.2 shows this pattern also holds when the model is simulated and the same regressions estimated on model-generated data: firms with high default risk see relatively larger increases in borrowing costs and relatively less access to new external finance than low-risk firms after an expansionary shock (Figure 6), exactly as in the data.

Heterogeneous firms invest in capital financed by internal funds or external (defaultable) borrowing, generating an external finance premium, embedded in a benchmark New Keynesian environment with sticky-price “retailer” firms generating a standard NK Phillips curve (Introduction, p. 2474-2475; Section 3). The model builds on Khan, Senga, and Thomas (2016)’s flexible-price financial-shock framework, extended with capital-quality shocks and a time-varying capital price “in order to generate variation in lenders’ recovery value of capital, as in the financial accelerator literature” (Introduction, p. 2476).

Optimal investment sets the marginal cost of capital (net-of-interest-savings relative price of capital, plus borrowing-cost terms) equal to the marginal benefit (discounted expected return on capital, plus a covariance term and a default-probability term). For a high-net-worth firm, “the marginal cost curve is flat when capital accumulation can be financed without incurring default risk, but becomes upward sloping when the borrowing required… creates default risk”; for a low-net-worth firm, this upward-sloping region begins at a lower level of capital, “since this firm has low initial net worth… it needs to borrow more than the risk-free firm to achieve the same level of investment” (Section 4, pp. 2487-2488).

For both firm types, the marginal benefit curve shifts out (lower discount rate, higher revenue product of capital from general-equilibrium price changes) and the marginal cost curve shifts up somewhat (higher relative price of capital q). For risky firms there are two additional effects on the cost curve: monetary policy raises net worth – via higher current cash flow, a higher value of undepreciated capital, and inflation-eroded real debt (eq. 11) – which “extends the flat region of the marginal cost curve,” and it raises lenders’ recovery value in default, which “flattens the upward-sloping region,” a channel the paper calls “reminiscent of the ‘financial accelerator’ in Bernanke, Gertler, and Gilchrist (1999)” (Section 4, pp. 2488-2489). Whether these effects are large enough to make risky firms more or less responsive than risk-free firms “is theoretically ambiguous,” which is why the paper turns to a disciplined calibration (Section 4, p. 2489).

Simulating a panel of firms through the calibrated model and re-estimating the empirical specification, the model generates a differential leverage response of about -1.30 to -1.47 (standardized), compared to -0.57 in the data – “just outside the 95% confidence interval of the empirical estimate” – with the model’s dynamics of the differential response over time “mostly stay[ing] within the data’s 90% confidence interval up to 8 quarters after the shock” (Section 6.2, Table VII, Figure 5). The authors note the model’s R² is higher than the data’s, “indicating that the data contain more unexplained variation than the model” – an explicit scope caveat on how much of the cross-sectional variation the model is meant to explain.

Feeding in only the path of the real interest rate (“direct effect”), only the relative price of capital, or only the remaining prices (output price, wages, inflation) separately, the paper finds that “firms with high default risk face a steeper marginal cost curve for financing investment” under all three channels, so heterogeneous responses are not attributable to any single price (Section 6.2, Figure 7). Critically, “both the direct and indirect effects of monetary policy play a quantitatively important role in driving the investment channel,” which “contrasts with Auclert’s (2019) and Kaplan, Moll, and Violante’s (2018) decomposition of the consumption channel,” where the direct interest-rate effect is small relative to indirect labor-income effects; the paper attributes this to firms being “more price-sensitive than households,” who are dampened by consumption-smoothing motives (Section 6.2, p. 2499).

Fixing firms’ state-contingent responsiveness but varying the initial cross-sectional distribution of net worth, “the average response of capital accumulation is 33% smaller starting from the low net-worth distribution… than starting from the steady state distribution,” where the low-net-worth reference distribution has “52% lower” average net worth and “37% more risky constrained firms” (Section 6.3, Table VIII). The authors interpret this as “a potentially powerful source of time-variation in the aggregate transmission mechanism: monetary policy is less powerful when net worth is low and default risk is high,” while explicitly flagging that the exercise varies the distribution exogenously and “has not been validated using aggregate time-series evidence” (Section 6.3, p. 2500).

Removing the nonnegativity constraint on dividends collapses the investment block to a financially unconstrained representative firm; comparing impulse responses, “the impact effect of monetary policy on investment is larger in our full model than in the representative firm benchmark… despite the fact that risky constrained firms are less responsive than risk-free constrained firms, both types of constrained firms are more responsive than in a model without financial frictions because expansionary monetary policy increases firms’ net worth” (Section 6.3, Figure 8, p. 2500). This reproduces, in a model with rich firm heterogeneity, the amplification result of Bernanke, Gertler, and Gilchrist (1999) – but the paper’s cross-sectional finding (low-risk firms are more responsive than high-risk firms) is a distinct, additional result that BGG’s representative-firm framework cannot speak to, since “movements in the marginal cost curve have a stronger effect” precisely because BGG’s constant-returns production implies a horizontal marginal benefit curve (Section 4, p. 2490).