Numerous central banks started to build the countercyclical capital buffer as bank profitability began to soar during the recent tightening cycle. Recent evidence suggests that building the buffer when there is headroom for doing so does not harm lending in the short-term and tends to increase it at longer horizons. This column proposes a quantitative macro-banking model that captures such evidence and illustrates how it can be applied to calibrate the positive neutral countercyclical capital buffer. Optimal dynamic capital buffers build in response to expected upward shifts in bank net interest margins. In doing so, they mitigate externalities due to bank risk failure and collateral constraints.
Supported by evidence (e.g. Jiménez et al. 2017, Couaillier et al. 2022, Bergant and Forbes 2023, Dursun-de-Neef et al. 2023, Mathur et al. 2023) and theory (e.g. Elenev et al. 2021, Corbae and D’Erasmo 2021), there is a wide consensus that the countercyclical capital buffer (CCyB) should be released in extraordinary crisis times. In contrast, the question of when and how to build such a buffer in normal times has been subject to intense debate in recent years. The CCyB was introduced in 2016 in the context of the Basel III Accords as a releasable capital buffer whose adjustments over the cycle were expected to take the credit-to-GDP gap as a key common reference. However, a large number of central banks started to build their releasable capital buffers during the recent tightening cycle as net interest margins began to soar and despite the fact that these and other contractionary shocks were exerting a downward pressure on aggregate demand and credit gaps were in negative territory (Figure 1). Recent evidence finds that building the CCyB when banks have headroom for doing so has no significant (negative) impact on short-term lending and increases lending at longer horizons through improved bank resilience (e.g. Bedayo and Galán 2024).
Figure 1 Net interest margin and credit-to-GDP gap in the euro area
Source: Muñoz and Smets (2025).
Notes: Bank net interest margin in the euro area computed as the difference between the non-financial corporations (NFC) loans interest rate and the household deposits interest rate and expressed in annualised percentage points. Credit-to-GDP in the euro area (secondary y-axis), constructed by the BIS (using the standard methodology made available on its webpage) and expressed as a percentage of GDP. The horizontal red line indicates the level below which the gap is negative. Data sources: European Central Bank (MFI Interest Rate Statistics) and BIS statistics.
Standard business-cycle macro-banking models that provide a convincing rationale for prudential bank capital regulation by featuring bank risk failure, limited liability, and deposit insurance generally find that there are little to no stabilisation and welfare gains from having a dynamic capital buffer akin to the CCyB when the optimal static capital requirement is already in place (e.g. Canzoneri et al. 2021, Abad et al. 2024). This is so as the externality they feature (i.e. agents do not internalise the impact their individual decisions have on the aggregate economic cost of bank risk failure) is most effectively corrected with fixed capital requirements. The very few exceptions to this result typically find that optimal dynamic capital requirements should respond to macroeconomic indicators such as credit and output (e.g. Davydiuk 2017).
In our new research (Muñoz and Smets 2025), we build a simple version of this class of models that explains why gradually building a dynamic capital buffer akin to the CCyB when expected net interest margins are increasing (i.e. when there is headroom for doing so) is optimal even when credit gaps are in negative territory.
The model: Externalities and bank capital requirements
We build a quantitative general equilibrium saver-borrower type of model that features collateral constraints as in Kiyotaki and Moore (1997), property markets à la Iacoviello (2005), and a banking sector as in Mendicino et al. (2020). The model features two frictions, each of which generates a different externality. First, individual banks default with some probability as they are hit by idiosyncratic asset return shocks (Bernanke et al. 1999) that capture any exogenous sources which may affect banks’ profitability. A fraction of bank deposits is uninsured and depositors price them based on the expected potential losses associated with the average bank risk failure. This friction implies that the risk of bank default is not priced at the margin and banks have incentives to lever up excessively (i.e. the capital adequacy constraint is binding) and to underprice the risk involved in lending to non-financial corporations. They do not internalise the impact their individual decisions have on the aggregate economic cost of bank risk failure. Second, borrowers (entrepreneurs) discount the future more heavily than savers (households), which effectively implies that property collateral constraints faced by entrepreneurs are binding in a neighbourhood of the steady state. As property prices enter these borrowing limits, this involves the presence of pecuniary externalities due to collateral constraints. Borrowers do not internalise the consequences their individual decisions have for these asset prices, which leads to financial amplification.
The only policy tool the competent authority has at hand to potentially correct these externalities is bank capital requirements. These requirements are set according to a simple policy rule that comprises a steady state component (i.e. static capital requirements) and a time-varying component that adjusts with a macro-financial indicator of the choice of the regulator (i.e. dynamic capital buffer).
Main findings
1. Optimal dynamic capital buffers are built in response to expected upward shifts in bank net interest margins (also referred to as the bank lending spread).
This result follows from the bank no-default condition, according to which the threshold for idiosyncratic bank asset return shocks below which banks default is fully determined by capital requirements and the components of forward-looking net interest margins. Other bank profitability metrics such as the bank return on equity (RoE) and return on assets (RoA) also enable dynamic capital buffers to induce significant stabilisation and welfare gains to the extent that they are highly and positively correlated with net interest margins and are informative of the bank profit generation capacity. In contrast, dynamic capital buffers guided by macroeconomic indicators are ineffective. Importantly, the calibrated model captures the positive but low historical correlation between net interest margins and macroeconomic indicators such as bank lending and the credit-to-GDP gap.
2. Optimal dynamic capital requirements mitigate externalities due to bank risk failure and collateral constraints over the cycle. The ‘collateral channel’ activates the full transmission mechanism through which the effects of accumulating capital buffers transmit to the real economy.
The above no-default condition and the externalities due to bank risk failure apply to all quantitative macro-banking models of this class. However, these dynamic capital buffers (guided by net interest margins) are often ineffective. By allowing for pecuniary externalities due to empirically relevant collateral constraints, what we refer to as the ‘collateral channel’ activates the full transmission mechanism through which the effects of accumulating releasable capital buffers in response to expected upward shifts in the bank lending spread transmit to the real economy. Figure 2 captures this transmission mechanism by displaying the impulse responses of selected aggregates to a negative financial (collateral) shock. The accumulation of releasable capital buffers when there is headroom for doing so (i.e. when the bank lending spread is expected to increase) strengthens banks’ resilience (i.e. the average bank default probability recedes). Through this mechanism, such policy action lowers bank resolution costs and taxes levied on households to fund the deposit insurance scheme. This frees up resources that help sustain aggregate demand. The build-up of capital buffers in this fashion: (i) has no significant negative impact on short-term lending while it increases lending at longer horizons through improved banks’ resilience, and (ii) mitigates pecuniary externalities due to collateral constraints (as revealed by the significant smoothing effect on the prices of the asset pledged as collateral).
Figure 2 Impulse responses to a negative financial (collateral) shock: Dynamic capital buffers
Source: Muñoz and Smets (2025).
Notes: Variables are expressed in percentage deviations from the steady state with the exceptions of bank capital requirements, the lending spread, and the bank default probability, which are shown as absolute deviations from the steady state and are expressed in percentage points, with all of them except for the first one being annualised. The solid line refers to the baseline calibration scenario. The dashed line refers to an alternative scenario that differs from the baseline one in that the CCyB parameter is set to a value of 0.2, with the indicator that guides CCyB decisions being the forward-looking growth rate of the bank lending spread. The size of the shock is set to 0.01.
In fact, the presence of these pecuniary externalities allows for dynamic capital buffers to mitigate the two types of externalities by managing banks’ resilience over the cycle. In their presence, there are mutually beneficial gains from trade between savers and borrowers. Entrepreneurs would like to borrow more to increase their consumption and investment. As the capital buffer builds in response to expected improvements in net interest margins, the average bank default probability recedes, causing: (i) the deposit rate (and the risk-free rate) to soar – encouraging households to save more – and the externality due to bank risk failure to be corrected as the excess return on bank deposits (over the risk-free rate) declines, (ii) aggregate demand to increase as bank resolution costs and taxes collected to fund the deposit insurance scheme fall (‘aggregate economic cost channel’). Higher household savings in deposits boost lending supply to entrepreneurs, who increase their consumption and property investments, thereby smoothing property prices and reducing the excess return on the asset they pledge as collateral (‘collateral channel’).
3. Optimal rules feature releasable capital buffers that are built very gradually (i.e. high persistence) and relatively lower static capital requirement levels (i.e. lower than those which are optimal in the absence of capital buffers).
Static capital requirements and releasable capital buffers provide stability to the system through different mechanisms. While the former permits effective correction of the externality due to bank risk failure in the steady state by managing the long-run level of bank default risk, the latter allows correcting externalities due to bank risk failure and collateral constraints by managing banks’ resilience over the cycle. That is, there are complementarities between the two but also a trade-off. In particular, the effectiveness of dynamic capital buffers decreases with the initial level of static capital requirements (i.e. increases with the steady-state average bank default probability). In fact, there is a lower bound for the long-run average bank default probability below which dynamic capital buffers are ineffective as there is no further space for stabilising the economy by strengthening banks’ resilience via buffer accumulation. Building these capital buffers very gradually is optimal; it amplifies the welfare gains of dynamic capital buffers and materially improves the trade-off between such buffers and static capital requirements. Figure 3 displays the impulse responses of selected aggregates to the same negative collateral shock under different scenarios to illustrate how the stabilisation gains of optimal dynamic capital buffers depend on the levels of static capital requirements and capital buffer smoothing.
Figure 3 Impulse responses to a financial shock: The role of static capital requirements and capital buffer smoothing
Source: Muñoz and Smets (2025).
Notes: Variables are expressed in percentage deviations from the steady state with the exceptions of bank capital requirements, the lending spread, and the bank default probability, which are shown as absolute deviations from the steady state and are expressed in percentage points, with all of them except for the first one being annualised. The green solid line refers to the baseline (calibration) scenario. The black dashed line refers to a scenario under which social welfare is maximised with respect to the CCyB parameter. The red diamond line relates to the optimal rule (i.e. social welfare is maximised with respect to the three policy parameters of the capital requirement rule). The blue dotted line refers to the same rule but with the persistence parameter being set to a value of zero. The indicator that guides CCyB decisions is the forward-looking growth rate of the bank lending spread. The size of the shock is set to 0.01.
Application: Calibrating the positive neutral rate for the countercyclical capital buffer
As an application of the quantitative analysis, we present a simple framework for calibrating micro- and macro-prudential capital requirements – including the so-called ‘positive neutral CCyB’ (PN-CCyB) – to give a sense of how the optimal rules would map into capital requirements and buffers typically calibrated by regulators in practice. Under the optimal rule with buffer smoothing, the calibrated optimal PN-CCyB – and, more generally speaking, calibrated optimal dynamic capital buffers – is larger (1.84%) than without it (0.37%), mainly because the volatility of the underlying indicator (i.e. the expected growth rate of the net interest margin) is higher (Figure 4). The reason for this is twofold. First, due to lower static capital requirements that make aggregates more volatile (from the bank lending spread to private consumption). Second, more gradual responses to expected shifts in the net interest margin cause such variables themselves to fluctuate more than in the absence of buffer smoothing. Ours complements other recently proposed frameworks to calibrate the positive neutral CCyB with empirical models (e.g. De Nora et al. 2025).
Figure 4 Calibrated optimal capital requirements
Source: Muñoz and Smets (2025).
Notes: Calibrated optimal capital requirements and each of its components for the cases in which Problem (23) is solved for γx, ργ and γ (Smoothing: YES) as well as for γx and γ (Smoothing: NO) with indicator eXt being the expected growth rate of the bank lending spread. In particular, the figure reports the calibrated optimal: (i) microprudential capital requirements (dashed area), (ii) PN-CCyB (solid area), and (iii) maximum cyclical component of the CCyB (dotted area). Capital requirements are expressed in percentage points.
Conclusion
Taken together, recent relevant empirical and theoretical studies suggest that the CCyB should be built in normal times, depending on the headroom for doing so, and fully released in extraordinary crisis times.
Source : VOXeu
