Monetary policy models: lessons from the Eurozone crisis

The paper addresses why, in the Eurozone after the 2007–2008 global financial crisis, prolonged expansionary monetary and fiscal policies coincided with deflationary pressures, falling credit volumes, and weak real activity. Standard frameworks (quantity theory, Keynesian and New-Keynesian models, limited participation and inside money models) cannot jointly reconcile expansionary policy with deflation and output contraction. Empirical evidence shows ECB’s stable inflation targeting, weak responsiveness of credit to money supply increases and interest rate cuts, and alterations in the monetary transmission mechanism with a prominent role of the bank lending channel and debt-deflation dynamics. The authors hypothesize that financial sector decisions—especially leverage (the loan-to-deposit ratio)—within the bank credit channel are central to a debt-deflation spiral that decouples policy from expected outcomes. They develop a dynamic general equilibrium model highlighting how ECB instruments interact with debt-deflation channels and then test the model’s long- and short-run implications using a VECM for the Eurozone.

The paper surveys theoretical and empirical strands relevant to Eurozone monetary transmission during the GFC. Quantitative theory can explain low inflation only under specific conditions (e.g., output gaps or collapsing velocity) not observed 2008–2018. Keynesian/New-Keynesian models—relying on a negative loan-volume/interest-rate relationship and expansionary policy raising output—do not fit Eurozone evidence. Inside money models can dampen inflationary effects via substitution of outside/inside money but still predict real expansion from policy not borne out in data. Empirical works highlight ECB’s inflation targeting, breakdowns or changes in transmission mechanisms, the role of the bank lending channel, and debt-deflation processes (e.g., Fisher, Minsky, Bernanke; Angeloni et al., ECB 2010; Fiedler & Gern 2019; Giannone et al. 2019). Deleidi (2018) finds an ambiguous relationship between interest rates and total loan aggregates, with other credit-condition variables mattering. Holtemöller (2004) motivates VEC approaches to capture long-run relations. Literature on financial leverage and balance-sheet channels (den Haan et al., Adrian & Shin, Schularick & Taylor, Claessens et al., Gambetti & Musso) underpins the focus on leverage. Macroprudential policy literature (Aiyar et al., Cerutti et al., IMF 2013) informs short-run credit effects.

Theoretical model: An extended limited participation dynamic general equilibrium model features households, firms, competitive financial intermediaries, and a central bank. Financial intermediaries own capital and channel deposits into productive capital subject to reserve requirements and money supply operations; they face a zero-profit condition. ECB instruments are: (i) open market operations (money supply, g_t), (ii) standing facilities setting an interest-rate corridor width m_t (deposit and marginal lending facility), and (iii) minimum reserve requirement s_t. The key state variable is the financial leverage (loan-to-deposit) ratio y_t = (1 − s_t + p_t g_t), where p_t is the endogenously determined fraction of money injections becoming productive capital, pinned down by the competitive zero-profit condition (1 − s_t + p_t g_t) r_t^P − r_t^B = 0 and the corridor m_t = r_t^P − r_t^B − δ_t. The model is recast as an equivalent social planner problem with cash-in-advance constraints; real allocations coincide with the competitive equilibrium. In steady state with constant s, g, m, the model yields long-run relations linking the real interest rate r and inflation π to leverage y. Specifically: r decreases with y (Eq. 8) and π increases with y (Eq. 9). The framework admits debt-deflation dynamics: deleveraging (falling y) reduces loans and aggregate demand, increases real rates, and depresses prices and activity, consistent with Fisher-Minsky-Bernanke channels. Empirical model: A structural VECM is estimated for monthly Eurozone data from November 2007 to December 2019 (ECB Statistical Data Warehouse). Variables: log core HICP (price level), real bank lending rate (weighted household mortgage and NFC >1y lending rates adjusted by SPF expected inflation), and log private-sector loan-to-deposit ratio (households+NPISHs+NFCs). Unit root testing (ADF, KPSS) indicates I(1) behavior after first differencing. A VAR lag length of four is selected by AIC. Cointegration is tested via Johansen with unrestricted intercepts and restricted linear trends in the cointegration space to accommodate differing deterministic trends. Two exogenous dummies are included: European sovereign debt crisis (Sep 2010–Feb 2012) and zero lower bound (from Mar 2016). Two cointegrating relationships are imposed per theoretical prediction, yielding error-correction terms used in the VEC equations. Impulse response functions (IRFs) are identified with Cholesky decomposition assuming recursive timing: price index ordered first; real interest rate innovations do not contemporaneously move prices; leverage shocks affect other variables with a one-month delay.

Theoretical: In steady state, there are two long-run relations among leverage (y), real interest rate (r), and inflation (π): r is negatively related to y (Eq. 8), and π is positively related to y (Eq. 9). Hence, ECB policy instruments (s, g, m) affect inflation and interest rates through y, highlighting the bank lending channel and consistency with debt-deflation mechanisms. Cointegration (2007–2019): Johansen tests indicate two cointegrating vectors between log core HICP, real interest rate, and log loan-to-deposit ratio, corroborating the theoretical long-run relations. Estimated cointegration equations (Table 2; Eqs. 10–11):

• π_t = −4.38042 + 0.00102·t + 0.03922·y_t
• y_t = 0.74244 − 0.002841·t − 0.01359·π_t Error-correction (Table 3): Significant ECT loadings indicate long-run adjustments led primarily by prices and leverage, not by the real interest rate:
• Δπ_t: η¹ (ECT1) = −0.08093*** (≈8% of deviation corrected monthly)
• Δy_t: η³ (ECT1) = −0.30748*** (≈31% correction), μ³ (ECT2) = −0.11187*** (≈11% correction)
• Δr_t: ECT loadings insignificant Impulse responses (Fig. 4):
• A positive real interest rate shock leads to a persistent decline in leverage, significant after ~3 months and lasting over 18 months (deleverage channel).
• A leverage shock raises the price level with significance from ~month 5 for about half a year and has a self-reinforcing effect on leverage (consistent with deleveraging/credit feedbacks and macroprudential tightening short-run effects).
• Inflation shocks have persistent positive effects on inflation itself and a simultaneous positive effect on leverage; longer-run dynamics reflect the negative r–π link via debt-deflation elements. Extended interval (1999–2020): Johansen tests suggest one cointegrating vector at conventional levels (Trace=52.21*** for r=0; r≤1 not significant), but two vectors are imposed for consistency. Cointegrating relations:
• π_t = 4.33029 + 0.00102·t + 0.18004·y_t
• y_t = −2.66626 + 0.01938·t − 0.75793·π_t ECT loadings (Table 5): Long-run adjustment shifts toward prices and real interest rates (η¹ for Δπ_t = −0.03549***; μ¹ for Δπ_t = −0.09399***), while leverage’s adjustment becomes insignificant. IRFs (Fig. 5): Leverage shocks provoke stronger positive responses in real interest rates (greater monetary policy effectiveness pre-crisis); real-rate shocks do not reduce leverage in the extended sample; leverage’s response to price shocks turns positive. These differences reflect structural regime changes (pre-crisis overheating/housing boom vs. crisis deleveraging and macroprudential tightening). Policy mechanisms and quantitative elements: The model implies that narrowing the standing facilities corridor (reducing m) increases y (since y = 1/(1 + A·m), A>0), supporting ECB measures such as reducing corridor width and introducing negative deposit facility rates (from June 2014). Empirically, 2014–2018 coincides with rising GDP growth and recovering loans to productive firms. Macroprudential tightening likely contributed to short-run credit contractions (lower p_t and y) despite long-run financial stability benefits.

Findings validate that Eurozone anomalies during 2008–2018—expansionary policy alongside deflation and weak activity—are consistent with a monetary transmission mechanism dominated by the bank lending channel and debt-deflation dynamics centered on leverage decisions. The theoretical DGEM shows that ECB instruments (reserves, money injections, corridor width) shape inflation and real rates through leverage (y). The VECM detects the predicted cointegrating relations and shows that deviations from long-run equilibria are corrected mainly by prices and leverage rather than real rates, confirming the central role of financial sector leverage decisions. IRFs reveal short-run dynamics shaped by crisis-specific factors: expansionary fiscal policies and sovereign risk episodes likely raised interest rates contemporaneously (crowding out), while macroprudential tightening compressed loan supply, reinforcing deleveraging. The extended-sample analysis shows regime-dependent transmission: pre-crisis overheating and feedback loan activity strengthen the interest rate response to leverage shocks and alter leverage’s responses, indicating structural changes and potential nonlinearities. Overall, the results integrate debt-deflation channels into a coherent transmission mechanism, explaining the coexistence of policy expansion with credit contraction and deflation.

The paper contributes a stylized monetary general equilibrium framework that embeds financial leverage into the transmission mechanism, generating long-run links between inflation, real interest rates, and bank lending consistent with ECB inflation targeting and debt-deflation theory. Empirically, a VECM for the Eurozone confirms two long-run relations during 2007–2019 and documents short-run dynamics consistent with deleveraging and policy interactions; an extended 1999–2020 analysis reveals regime-specific differences in adjustments and IRFs. Policy implications include: (i) acting directly on leverage by narrowing the standing facilities corridor (lower m) and maintaining negative deposit facility rates to raise the fraction of money injections reaching productive firms (higher p_t) and thus y; (ii) considering reductions in reserve requirements (s) and increases in money supply (g), including asset purchases, while recognizing limits if leverage is falling; (iii) advancing structural financial policies (banking union, competition with stability) to improve the lending channel; (iv) potentially revising inflation objectives upward and complementing with targeted government purchases under debt-deflation. The framework also motivates future research on modeling leverage’s microfoundations (bank/household leverage, LTVs, capital ratios) and on nonlinear, regime-dependent transmission mechanisms observed across crisis and non-crisis periods.

Short-run dynamics of the DGEM lack a closed-form solution and require numerical methods; the paper relies on VECM IRFs for short-run insights, which are sensitive to identification, lag selection, and crisis-specific regimes. The extended-interval cointegration rank is weaker (only one vector significant at conventional levels), indicating structural breaks/regime shifts that complicate uniform long-run relationships. The model abstracts from explicit optimal bank capital requirements, default probabilities, and heterogeneity (e.g., bank size), limiting direct policy calibration. Macroprudential policy effects are heterogeneous across tools and economies, making attribution in IRFs partial. Measurement choices (e.g., SPF-based expected inflation, loan-rate weighting, core HICP) may affect estimates.