Economic analysis of credit markets often relies on a model that assumes firms have a uniform demand for credit across all their lenders, but a new paper from the European Central Bank (ECB) has shown that this long-held assumption is empirically untenable.

The working paper, titled Identifying relationship-level effects using covariance restrictions and authored by Olivier De Jonghe and Daniel Lewis, proposes a new methodology to decompose credit outcomes into specific demand and supply shocks at the relationship level.

The research suggests that the dominant tool for separating these effects, known as the two-way fixed effects estimator, frequently masks substantial heterogeneity that is vital for understanding how economic policy impacts firms.

“The homogeneity assumption is both conceptually restrictive and, as we show, empirically untenable,” the authors note in the report.

Rather than assigning a single demand shock to each firm as if its needs were identical across all lenders, the authors demonstrate how to recover a separate demand and supply shock for every firm-bank pair in every period.

The study applied this framework to thousands of firms and banks across nine euro area countries using the AnaCredit dataset, spanning the pandemic, the post-pandemic inflation surge, and the subsequent monetary policy tightening.

When testing the homogeneity assumption, the researchers found that it failed in 25 of the 27 country-period combinations studied, confirming that conventional approaches often produce unreliable results.

A central finding of the research is that the standard approach conflates genuine firm demand with idiosyncratic bank supply conditions, which leads to significant statistical bias.

In a striking example, the paper illustrates that when measured using standard firm-time fixed effects, a positive demand shock is often incorrectly associated with a lower interest rate, which is the direct opposite of what economic theory predicts.

The researchers used the ECB’s 2022 monetary policy tightening as a quasi-natural experiment to demonstrate how this bias manifests in the real economy.

While the ECB raised rates by 450 basis points over fourteen months, firms borrowing at floating rates experienced a significant contraction in credit supply compared to those that had locked in fixed rates.

Conventional methods, which rely on the standard fixed effects approach, severely underestimated these effects, rendering the impact of the monetary contraction on firm activity nearly invisible.

However, the authors’ newly identified demand measure successfully restores a large and highly significant estimate of these effects.

The difference in measurement translates into real-world consequences, as the study documents that floating-rate borrowers experienced significantly lower total assets and turnover over the following two years.

Furthermore, the new methodology allows for a more granular view of how monetary policy shocks propagate through the financial system.

For instance, following contractionary policy surprises, firms with a higher share of fixed-rate loans reduce their demand for new credit, while banks simultaneously increase their supply to those same firms because they view them as safer counterparties.

The authors suggest that these findings have broad implications for empirical research far beyond corporate credit markets.

The methodology is general enough to be applied to labour markets, trade networks, and other areas of industrial organisation where identifying two-sided effects is crucial.

“We believe that empirical results of this nature are novel, and thus provide a new set of empirical facts to discipline theoretical models of the transmission of various types of policies,” the authors stated.

The research also suggests that a reevaluation of a large body of empirical work in finance and economics may be necessary, particularly in studies that rely on the classic two-way fixed effects paradigm.

By demonstrating that within-firm variation in credit demand is roughly one-third as large as variation across different firms, the paper underscores why treating all of a firm’s relationships as homogeneous can lead to misleading conclusions.

This study provides a new statistical toolkit that researchers can use to achieve more precise identification of structural shocks in complex, bipartite networks.

As policymakers continue to navigate the distributional effects of interest rate changes, the ability to accurately distinguish between credit supply disturbances and genuine demand shifts is becoming an increasingly important requirement for sound economic analysis.