The Transmission of Monetary Policy to Corporate Investment: the Role of Loan Renegotiation
What this paper finds — and why it matters
Layer 1: Overview
Research question and motivation. This paper asks how monetary policy transmits to corporate investment through bank credit, and specifically whether the relevant credit margin is the origination of new loans (the channel emphasized by the traditional credit/bank-lending channel literature, e.g., Kashyap, Stein and Wilcox, 1993) or the renegotiation of existing loans. The motivation is institutional: in the U.S., almost 70% of corporate loan contracts are renegotiated prior to maturity, with firms renegotiating existing loans about twice as often as issuing new ones, and renegotiations typically alter loan amounts, spreads and maturities by 30%–40% of initial values. Prior work measured only new lending, disregarding these revisions. The author claims this is the first study to distinguish new loans from revisions of existing loan terms in the transmission channel.
Data and empirical strategy. The author builds a novel loan-level panel by combining automated textual analysis with manual review of SEC EDGAR credit-agreement filings (2005–2015, spanning conventional and unconventional/ZLB policy). Each loan path is traced from origination through renegotiations to maturity/early termination. After standard restrictions the loan-level sample has 9,565 loan paths from 2,685 firms, totaling 129,733 loan-quarter observations; ~53% of observations are private firms. Dataset accuracy exceeds 94% versus Roberts (2015)’s hand-collected data (~90% of ~300 matched observations agree completely). Loan data are merged with Compustat, Call Report, DealScan, FISD/SDC. The impulse is the Bu, Rogers and Wu (2021) monetary policy shock series (covers conventional + unconventional policy, purged of information effects), aggregated to quarterly. Identification uses local projections (Jordà, 2005): a linear probability model at the bank-firm-quarter level for the extensive margin of credit (origination vs renegotiation indicator), an intensive-margin variant using cumulative standardized within-bank-firm demeaned loan amount/spread, and a firm-quarter investment-response regression. Shocks are normalized so positive = expansionary.
Main quantitative findings. A 25bps expansionary shock raises the renegotiation probability by about 1.7–2.1 percentage points in the same quarter (economically large vs the ~10%, specifically 10.2%, average quarterly renegotiation rate), persisting for about three quarters. The effect on new-loan origination is positive but weaker and varies across specifications (~0.3–1.5 pp). On the intensive margin, renegotiation expands loan amount by ~0.2 standard deviations vs average renegotiations, with no significant spread increase; new-loan volume shows limited/weak evidence of increase (origination amount coefficient -0.184*, spread insignificant). Effects are asymmetric: expansionary shocks matter more than contractionary ones on the extensive margin (Wald test rejects symmetry for renegotiation p=0.000 and origination p=0.013), but not the intensive margin. For investment: firms that renegotiate raise investment relatively more than non-renegotiators, with the relative effect notable from 3 quarters and peaking at 10 quarters—faster than the average response, which peaks at 18 quarters (where a 25bps expansionary shock raises the investment rate up to ~0.2%). Heterogeneity: highly leveraged & bank-dependent firms have ~3–4 pp higher origination/renegotiation propensity after the shock, and renegotiation amplifies their investment response. New-loan issuance, by contrast, is driven by prior investment growth (firms with prior investment/assets one SD above average are ~0.7 pp more likely to originate). Contribution to the aggregate: renegotiating firms account for ~47.4% [43.6, 51.4] of the average investment response, originating firms ~11.9% [8.5, 15.2], and either activity ~55.1% [51.3, 58.8].
Implications. Renegotiation, not new origination, is the dominant bank-credit channel transmitting monetary policy to investment, it acts faster than origination, and it amplifies responses for financially constrained firms—implying monetary policy eases their constraints via improved credit access through renegotiation. Policymakers should monitor renegotiation dynamics, not just total loan balances, and coordinate prudential and monetary policy since prudential regulation affects renegotiation conditions.
Layer 2: Deep Dive
What is the identification strategy and what are the main threats to it?
The author uses local projections (Jordà, 2005) with the Bu, Rogers and Wu (2021) monetary policy shock series as the exogenous impulse. That shock is constructed to be exogenous (heteroskedasticity-based partial least squares isolating monetary from non-monetary news), purged of central-bank information effects, and largely unpredictable from Blue Chip forecasts/news/sentiment, addressing the standard confounding of policy actions with the central bank’s economic outlook. For the credit-margin regressions, bank and firm fixed effects (and in saturated specs, bank-by-firm fixed effects) absorb persistent supply- and demand-side and relationship heterogeneity; in the heterogeneity regressions bank-by-time fixed effects absorb credit-supply variation so the interaction identifies demand-side variation. Standard errors are two-way clustered. Threats: generated-regressor inference (the shock is estimated), which the author notes Pagan (1984) shows yields consistent SEs under the null and which holds when using shocks as instruments for interest rates; and demand-supply confounding, addressed via fixed effects. A subtler concern is reverse selection in investment regressions—firms renegotiating because investment is already trending up—which the paper addresses head-on in the decomposition (Section 3.2.3).
What are the main mechanisms and how are they distinguished empirically?
The core distinction is renegotiation vs new origination. Renegotiation responds strongly and immediately to expansionary shocks (1.7–2.1 pp), expands borrowing (~0.2 SD) without raising spreads, and is independent of prior investment growth. Origination responds weakly, and its likelihood is instead predicted by the firm’s prior investment growth (~0.7 pp per SD), so it follows rather than drives investment. The decomposition (Table 8) separates total discounted investment growth (t-1 to t+18) into ’lead’ (t to t+18) and ’lagged’ (t-1 to t) components: for renegotiating firms the total response (0.537**) is driven by the lead component (0.707***) not the lagged (-0.178, insignificant), confirming renegotiation predicts subsequent investment; for originating firms none of total/lead/lagged is significant. The paper also reasons that renegotiation is cheaper (fee ~0.1–0.3% of loan vs origination fee ~0.5–5% plus search/matching costs) and yields a larger borrower surplus, explaining why firms prefer it after accommodative shocks.
What heterogeneity is documented?
(1) By financial constraint: highly leveraged & bank-dependent firms (15.8% of firm-quarter obs) show ~3–4 pp higher semi-elasticity of both origination and renegotiation propensity after a 25bps expansionary shock, and renegotiation significantly magnifies their investment response (triple-interaction, Figure 5). (2) By prior investment: firms with high ex-ante investment growth are more likely to originate (not renegotiate). (3) By age: younger firms rely more on new-loan issuance than renegotiation. (4) Alternative constraint proxies (size, leverage, distance to default, younger-and-non-dividend) in appendix figures confirm constrained/closer-to-default firms have higher credit-adjustment likelihood. (5) By renegotiation subtype: amount, spread and covenant adjustments produce greater relative investment responses, but maturity changes do not. Notably the intensive-margin loan-amount response shows NO significant heterogeneity by constraint or prior investment (Table 6).
What robustness checks are run?
Controlling for lender-specific bank capital ratio (Table B.1.1); estimating at the more granular loan-quarter level (Table B.1.2); an alternative construction of zeros for the origination indicator covering all ever-matched bank-firm pairs (Table B.1.3, which shows no immediate origination effect but lagged effects—widening the renegotiation/origination gap); using central-bank information shocks of Jarociński and Karadi (2020), which have the opposite sign on credit propensity, consistent with the information-effect interpretation (Table B.1.4); using the shock as an instrument for interest-rate changes (results unchanged); alternative shock series (Nakamura-Steinsson; Jarociński-Karadi); a nonlinear (logit/probit) procedure; and an alternative unweighted quarterly shock aggregation. The micro data also reproduce macro investment dynamics (~0.9 correlation with BEA private nonresidential fixed investment), validating external relevance.
How does this paper relate to and differ from closely related prior work?
It extends the bank-lending and firm-balance-sheet credit-channel literature (Kashyap-Stein-Wilcox 1993; Jiménez et al. 2012; Abuka et al. 2019) which measured only new lending, by separating renegotiation. It extends Ippolito, Ozdagli and Perez-Orive (2018)’s floating-rate channel by showing renegotiation alters loan terms in ways that can dominate the mechanical floating-rate/policy-rate link. It vastly expands the renegotiation data of Roberts (2015) (114 firms) and Roberts and Sufi (2009) via text mining, and is more comprehensive than supervisory SNC/Y-14 data (which miss major renegotiation types). On heterogeneity it complements Caglio, Darst and Kalemli-Özcan (2021), Jeenas (2019), Ottonello and Winberry (2020), and Cloyne et al. (2023). On asymmetry it aligns with Kandil (1995) and extends Abuka et al. (2019) (asymmetry on extensive but not intensive margin). It links to Lummer and McConnell (1989) on the informational distinctness of renegotiated vs new loans, and to Mian and Santos (2018) on renegotiation and capex over the credit cycle.
What are the policy implications and their scope conditions?
Because monetary policy transmits to investment with a lag while renegotiation responds immediately, renegotiation can serve as an early predictor of effective transmission, so policymakers should monitor renegotiation dynamics—not just total loan balances. Renegotiation is described as potentially ’the sole lifeline’ for financially constrained firms, magnifying their investment response. The paper highlights coordination between micro/macroprudential policy and monetary policy, since prudential regulation affects renegotiation lending conditions (Thakor and Furlong Wilson, 1995); depending on objectives, regulators might relax or tighten renegotiation conditions. Scope conditions: estimates apply to U.S. firms 2005–2015 spanning conventional and unconventional/ZLB regimes; effects are stronger for expansionary than contractionary shocks (asymmetry); and the author flags that the renegotiation channel’s role may differ between conventional and unconventional periods as a topic for future research.
What significant caveats or measurement details apply?
Renegotiations bundle amendments, amended-and-restated agreements and replacements, recorded together because the economic distinction is minor (following Roberts, 2015). Pre-specified contractual changes (rating-triggered spread increments, Evergreen auto-extensions) are NOT counted as renegotiations. Loans are assumed matured absent contrary SEC evidence. Intensive-margin samples are much smaller (conditional on the event and on non-missing spreads). The firm-quarter investment sample requires firms observed at least 6 years (24 quarters). Observations with negative bank capital (<0.4%, mostly during the GFC) are excluded. Balance-sheet variables are winsorized at 1% (0.5% for some). The investment-rate mean is ~0.2 (capxq*4/lagged ppentq); average bank capital ratio is 12.2% (SD 4.8%).