RESEARCH ARTICLE研究文章

Bank Capital Requirements, Lending Supply, and Economic Activity: A Scenario Analysis Perspective
银行资本要求、贷款供应与经济活动：情景分析视角

Antonio M. Conti $^{1}$ | Andrea Nobili $^{2}$ | Federico M. Signoretti $^{1}$ $^{1}$ Directorate General for Economics, Statistics and Research, Banca d'Italia, Roma, Italy | $^{2}$ Directorate General for Currency Circulation and Retail Payments, Banca d'Italia, Roma, Italy
$^{1}$ 意大利银行经济、统计与研究总局，罗马，意大利 | $^{2}$ 意大利银行货币流通与零售支付总局，罗马，意大利Correspondence: Antonio M. Conti (antoniomaria.conti@bancaditalia.it)
通信：Antonio M. Conti (antoniomaria.conti@bancaditalia.it)Received: 5 November 2024 | Accepted: 6 November 2024
收到：2024 年 11 月 5 日 | 接受：2024 年 11 月 6 日

Abstract摘要

We evaluate the relation between bank capital, lending supply, and economic activity using Italian data over 1993-2015, a period which covers three key post-crisis regulatory and supervisory measures (the Basel III reform, the 2011 European Banking Authority [EBA] stress test, the European Central Bank’s [ECB] Comprehensive Assessment, and launch of the Single Supervisory Mechanism-SSM). We quantify the impact of increased bank capital requirements using a novel procedure that recovers the magnitude of the policy measures, relying on scenario analysis and Bayesian VARs with a rich characterization of the banking sector. We document that the EBA and SSM measures unpredictably raised Tier 1 ratio by about 2.5 percentage points, leading to an average reduction in credit to firms and households by $5 %$ and $4 %$ , respectively, and to a decline in real GDP by over $2 %$ and $4 %$ . The Basel III bank capital increase is instead correctly anticipated in out-of-sample forecasting. These findings are robust to time-varying model parameters and consistent with narrative sign restriction techniques.
我们使用 1993 年至 2015 年的意大利数据评估银行资本、贷款供应和经济活动之间的关系，这一时期涵盖了三项关键的后危机监管和监督措施（巴塞尔 III 改革、2011 年欧洲银行管理局[EBA]压力测试、欧洲中央银行[ECB]综合评估以及单一监督机制-SSM 的启动）。我们使用一种新颖的程序量化提高银行资本要求的影响，该程序通过情景分析和具有丰富银行业特征的贝叶斯 VAR 恢复政策措施的规模。我们记录到 EBA 和 SSM 措施不可预测地将一级资本比率提高了约 2.5 个百分点，导致对企业和家庭的信贷平均减少了 $5 %$ 和 $4 %$ ，并导致实际 GDP 下降超过 $2 %$ 和 $4 %$ 。而巴塞尔 III 银行资本的增加则在样本外预测中被正确预期。这些发现对时间变化的模型参数是稳健的，并与叙述性符号限制技术一致。

JEL Classification: C11, C32, C45, E32, F34, G38
JEL 分类：C11, C32, C45, E32, F34, G38

1 | Introduction1 | 引言

A large body of theoretical and empirical research has investigated how banking regulators and supervisors’ requests of higher bank capitalization affect the willingness of intermediaries to extend credit and, in turn, economic activity. While a stronger capital base has several financial stability benefits in the long-run, the move towards higher levels of capitalization could be accompanied by credit supply restrictions in the shortrun. In particular, banks may decide to meet the requirements to increase their ratios by reducing their exposures to customers and/or by charging higher lending rates, reflecting the higher cost of equity relative to other sources of funding. In response to the deterioration in borrowing conditions, households and firms may reduce or postpone their spending and investment plans.
大量的理论和实证研究探讨了银行监管机构和监督者对更高银行资本化的要求如何影响中介机构扩展信贷的意愿，从而影响经济活动。虽然更强的资本基础在长期内具有多种金融稳定的好处，但向更高资本化水平的转变可能在短期内伴随信贷供应限制。特别是，银行可能会通过减少对客户的风险敞口和/或提高贷款利率来满足提高其比率的要求，这反映了相对于其他融资来源更高的股本成本。作为对借贷条件恶化的回应，家庭和企业可能会减少或推迟其支出和投资计划。

In this paper, we study the impact on economic activity and credit supply of regulatory and supervisory initiatives that resulted in increased pressure on banks to improve their capital ratios. Our empirical strategy consists in estimating a large Bayesian vector autoregression (BVAR) model on quarterly Italian data over the period 1993-2015 and then to use scenario analysis for three specific events in which the authorities increased pressure on banks to raise their capital levels. The Italian banking system is particularly well-suited to address this issue because it has been under close scrutiny by supervisors due to the high level of legacy assets inherited from the European sovereign debt crisis and because it has been affected-along with the other European countries that have joined the banking union-by the important changes in regulatory and supervisory policy associated with the introduction of the Single Supervisory Mechanism (SSM).
在本文中，我们研究了监管和监督措施对经济活动和信贷供应的影响，这些措施导致银行面临更大的压力以改善其资本比率。我们的实证策略包括在 1993 年至 2015 年期间对季度意大利数据估计一个大型贝叶斯向量自回归（BVAR）模型，然后对三个特定事件进行情景分析，在这些事件中，监管机构加大了对银行提高资本水平的压力。意大利银行系统特别适合解决这个问题，因为由于从欧洲主权债务危机中继承的高水平遗留资产，它一直受到监管机构的密切审查，并且由于与其他加入银行联盟的欧洲国家一起，它受到与引入单一监管机制（SSM）相关的监管和监督政策的重要变化的影响。

Our contribution to the literature is twofold. The first one is to assess and quantify the impact of the main regulatory and supervisory initiatives that have raised pressures on Italian banks to increase their capital levels since the onset of the Global Financial Crisis. Specifically, we focus on the evolution of the Tier 1 ratio, a key regulatory measure of a bank’s capital adequacy, defined as the ratio between Tier 1 capital and riskweighted assets (RWAs) (total bank assets weighted according to their riskiness). While some papers have documented the existence of negative short-run effects of increases in capital requirements, there is no consensus on the magnitude of these effects.

^{1}

Moreover, we use a broad definition of regulatory/supervisory initiatives, that includes, but is not limited to, increases in minimum capital requirements, whether broad-based (e.g., “Pillar 1” requirements; see below) or bank-specific (e.g., Pillar 2). Indeed, we also aim to capture the effect of increases in banks’ capital as a result of non-binding supervisory initiatives, such as Pillar 2 Guidance (P2G, i.e., the level of capital that the supervisor considers adequate to provide a sufficient buffer to withstand stressed situations) and increases in banks’ capital that banks have undertaken in anticipation of/to avoid shortfalls that may then trigger bank actions (such as the stress tests).
我们对文献的贡献有两个方面。第一个是评估和量化自全球金融危机开始以来，主要监管和监督举措对意大利银行提高资本水平施加的压力的影响。具体而言，我们关注一级资本充足率的演变，这是衡量银行资本充足性的关键监管指标，定义为一级资本与风险加权资产（RWAs）（根据风险程度加权的总银行资产）之间的比率。虽然一些论文记录了资本要求增加的短期负面影响的存在，但对这些影响的大小没有共识。此外，我们使用广泛的监管/监督举措定义，包括但不限于最低资本要求的增加，无论是广泛的（例如，“第一支柱”要求；见下文）还是特定于银行的（例如，第二支柱）。确实，我们还旨在捕捉由于非约束性监管举措（如第二支柱指导（P2G，即监管机构认为足够的资本水平，以提供足够的缓冲以应对压力情况））而导致的银行资本增加的影响，以及银行为预期/避免可能触发银行行动（如压力测试）的短缺而进行的资本增加。

Our second contribution, of a methodological nature, is to provide a macroeconomic approach for measuring the impact of regulatory and prudential pressures on bank capitalization and, consequently, on credit supply and economic activity. Moreover, our framework yields good forecasting properties for bank capital in times of crisis and possible structural breaks.
我们的第二个贡献是方法论性质的，提供了一种宏观经济方法来衡量监管和审慎压力对银行资本化的影响，从而影响信贷供应和经济活动。此外，我们的框架在危机时期和可能的结构性断裂中对银行资本具有良好的预测特性。

The literature has used several methodologies to study the impact of shocks to bank capital, all of which have some limitations. Some studies have relied on direct observation of bankspecific capital requirements (Meeks 2017; Aiyar, Calomiris, and Wieladek 2016; De Jonghe, Dewachter, and Ongena 2020; Raja 2023), which are often unavailable to researchers. Other researchers have used event studies or quasi-random experiments (Jiménez et al. 2017; Mésonnier and Monks 2015; Gropp et al. 2019; Galardo and Vacca 2022), which unfortunately are not well suited to studying the effects of policy measures that take some time to propagate and whose effects are hardly exhausted in a narrow time window (D’Amico 2016).

^{2}

Finally, other papers have estimated structural VARs with short-run restrictions, either sign or zero (Kanngiesser et al. 2020; Noss and Toffano 2016). While this is a popular and well-established method in the macroeconomic literature, it has some limitations when applied to the research question under consideration. under evaluation. For example, it is particularly difficult to impose convincing sign restrictions in VAR models larger than 8-10 variables, forcing researchers to restrict the analysis to key macroeconomic and credit variables

^{3}

; Moreover, sign restrictions often lead to shortlived and possibly on-impact peak effects of exogenous policy interventions, a feature plausibly at odds with feature plausibly at odds with the lags with which banks adjust their balance sheets.

^{4}

We address these and several other issues in a companion paper (Conti, Nobili, and Signoretti 2023), in which we identify shocks to bank capital requirements by adding narrative restrictions (Antolín-Díaz and Rubio-Ramírez 2018).
文献中使用了几种方法来研究银行资本冲击的影响，但这些方法都有一些局限性。一些研究依赖于对银行特定资本要求的直接观察（Meeks 2017；Aiyar, Calomiris, 和 Wieladek 2016；De Jonghe, Dewachter, 和 Ongena 2020；Raja 2023），这些要求通常对研究人员不可用。其他研究者使用了事件研究或准随机实验（Jiménez et al. 2017；Mésonnier 和 Monks 2015；Gropp et al. 2019；Galardo 和 Vacca 2022），但不幸的是，这些方法并不适合研究需要一些时间才能传播的政策措施的效果，并且其效果在狭窄的时间窗口内几乎无法耗尽（D’Amico 2016）。最后，其他论文估计了具有短期限制的结构 VAR，限制可以是符号或零（Kanngiesser et al. 2020；Noss 和 Toffano 2016）。虽然这在宏观经济文献中是一种流行且成熟的方法，但在应用于所考虑的研究问题时存在一些局限性。例如，在超过 8-10 个变量的 VAR 模型中施加令人信服的符号限制特别困难，这迫使研究人员将分析限制在关键的宏观经济和信贷变量上

^{3}

；此外，符号限制往往导致外生政策干预的短期和可能的即时峰值效应，这一特征可能与银行调整其资产负债表的滞后特征相悖。

^{4}

我们在一篇伴随论文中解决了这些及其他几个问题（Conti, Nobili, and Signoretti 2023），在其中我们通过添加叙述限制（Antolín-Díaz and Rubio-Ramírez 2018）来识别银行资本要求的冲击。

We use counterfactuals based on large Bayesian VAR models (BVAR) and computed over three short time windows following specific regulatory and/or supervisory events in which the authorities increased pressure on banks to increase their capital levels and at the same time observed large and persistent increases in the Tier 1 ratio. This approach mirrors the literature on monetary policy, where this methodology has been used to study the impact of unconventional monetary policy measures (Lenza, Pill, and Reichlin 2010; Giannone et al. 2012; Kapetanios et al. 2012; Altavilla, Giannone, and Lenza 2016; Dahlhaus, Hess, and Reza 2018; Altavilla, Canova, and Ciccarelli 2020).
我们使用基于大型贝叶斯 VAR 模型（BVAR）的反事实，并在特定监管和/或监督事件后计算三个短时间窗口，在这些事件中，监管机构加大了对银行提高资本水平的压力，同时观察到一级资本比率的大幅和持续增长。这种方法反映了货币政策的文献，其中这种方法已被用于研究非常规货币政策措施的影响（Lenza, Pill, and Reichlin 2010; Giannone et al. 2012; Kapetanios et al. 2012; Altavilla, Giannone, and Lenza 2016; Dahlhaus, Hess, and Reza 2018; Altavilla, Canova, and Ciccarelli 2020）。

The first step is to retrieve a counterfactual path for the Tier 1 capital ratio in each episode of regulatory and/or supervisory policy intervention. To do this, we follow two strategies, both based on out-of-sample forecasts, after estimating the model up to the last data point before the policy intervention. First, following Altavilla, Canova, and Ciccarelli (2020), we let the Tier 1 ratio evolve unconditionally according to its BVAR forecast.

^{5}

This counterfactual can be interpreted as the path of the Tier 1 ratio that might have been expected on the basis of historical regularities. In the second strategy, the counterfactual path of the Tier 1 ratio is the forecast of the Tier 1 ratio conditional on the expected path of the main determinants of bank capital in real time. This path is the one consistent with the latest available vintage of the Eurosystem’s macroeconomic projection exercise (MPE) prior to the end of the estimation sample.

^{6}

This counterfactual can be interpreted as the expected path of Tier 1 ratio based on the exogenous disturbances that were foreseeable in real time. The conditional forecast approach is similar to the counterfactual balance sheet exercises in Lenza, Pill, and Reichlin (2010) and Giannone et al. (2012). However, both papers compute their conditional forecast on realized rather than expected values of potential determinants of the policy variable (in their case the Eurosystem balance sheet, in our case the Tier 1 ratio). Once the counterfactual Tier 1 ratios are obtained, their difference from the actual developments can be interpreted as a measure of the size of the policy actions that took place during each time window.

^{7}

第一步是为每个监管和/或监督政策干预的情景检索一个反事实路径，以获取一级资本比率。为此，我们遵循两种策略，均基于样本外预测，在估计模型直到政策干预前的最后数据点后进行。首先，遵循 Altavilla、Canova 和 Ciccarelli（2020）的做法，我们让一级资本比率根据其 BVAR 预测无条件演变。

^{5}

这个反事实可以被解释为基于历史规律可能预期的一级资本比率路径。在第二种策略中，一级资本比率的反事实路径是基于实时主要银行资本决定因素的预期路径的一级资本比率预测。这个路径与估计样本结束前最新可用的欧洲系统宏观经济预测练习（MPE）的一次性版本一致。

^{6}

这个反事实可以被解释为基于实时可预见的外生干扰的一级资本比率的预期路径。条件预测方法类似于 Lenza、Pill 和 Reichlin（2010）以及 Giannone 等人（2012）中的反事实资产负债表练习。然而，这两篇论文计算的条件预测是基于政策变量的潜在决定因素的实现值，而不是预期值（在他们的案例中是欧元体系资产负债表，在我们的案例中是一级资本比率）。一旦获得反事实一级资本比率，它们与实际发展的差异可以被解释为在每个时间窗口内发生的政策行动的规模度量。

^{7}

As a second step in our analysis, we compare the evolution of lending volumes, lending spreads, GDP, and inflation in a simulation of the model conditional on the observed path of the Tier 1 ratio (policy scenario) and in one conditional on its counterfactual path (no-policy scenario). The two simulations differ in their assumptions about the evolution of bank capital but are otherwise identical. As in Giannone et al. (2012), we can thus (loosely) characterize this difference as an impulse response of each variable to the regulatory/supervisory actions.
作为我们分析的第二步，我们比较了在基于观察到的一级资本比率路径（政策情景）和基于其反事实路径（无政策情景）的模型模拟中，贷款量、贷款利差、GDP 和通货膨胀的演变。这两个模拟在银行资本演变的假设上有所不同，但在其他方面是相同的。因此，正如 Giannone 等人（2012）所述，我们可以（宽泛地）将这种差异描述为每个变量对监管/监督行动的脉冲响应。

Why can the difference between the actual and the counterfactual Tier 1 ratio be reasonably interpreted as the size of the policy actions? First, our interpretation is supported by a compelling narrative during the considered event windows, as the authorities explicitly asked Italian banks to significantly increase their capital levels.

^{8}

In more detail, the first time window starts in the second quarter of 2009 and coincides with the discussion on the reform of prudential regulation (Basel III reform), which aimed at improving the quantity and quality of capital and curbing excessive financial leverage in the aftermath of the Global Financial Crisis.

^{9}

In the considered period, total capital raised amounted to over 20 billion. The second period starts in the first quarter of 2011 and covers the European Banking Authority(EBA) 2011 stress test and capital exercise; during this period, banks raised about 25 billion in capital both in anticipation of the stress-test results and as a result of the additional buffers required by the Capital exercise.

^{10}

The third forecast period starts in the first quarter of 2014 and overlaps with the implementation of the ECB CA and the first months of the SSM. During this period, total capital increased by around 20 billion.

^{11}

Given this narrative and the magnitude of the authorities’ capital requests (amounting, overall, to about

1 / 3

of total banks’ equity in 2008), it is very unlikely that the difference between the observed and the counterfactual evolution of bank capital-which can be thought of as the most likely linear combination of shocks not captured by the large conditioning set that hit the economy over the narrow forecast window-is driven by factors unrelated to those policy actions.
为什么实际与反事实一级资本充足率之间的差异可以合理地解释为政策行动的规模？首先，我们的解释得到了在考虑的事件窗口期间的有力叙述的支持，因为当局明确要求意大利银行显著提高其资本水平。

^{8}

更详细地说，第一个时间窗口始于 2009 年第二季度，并与关于审慎监管改革（巴塞尔协议 III 改革）的讨论相吻合，该改革旨在改善资本的数量和质量，并在全球金融危机之后遏制过度的金融杠杆。

^{9}

在考虑的期间，总共筹集的资本超过 200 亿。第二个时期始于 2011 年第一季度，涵盖了欧洲银行管理局（EBA）2011 年压力测试和资本练习；在此期间，银行筹集了约 250 亿资本，既是为了预期压力测试结果，也是由于资本练习所需的额外缓冲。

^{10}

第三个预测期始于 2014 年第一季度，并与欧洲央行 CA 的实施以及 SSM 的前几个月重叠。在此期间，总资本增加了约 200 亿。

^{11}

鉴于这一叙述和当局资本请求的规模（总体约占 2008 年总银行股本的

1 / 3

），观察到的银行资本与反事实演变之间的差异——可以认为是未被在狭窄预测窗口内影响经济的大型条件集捕捉到的最可能的线性组合冲击——极不可能是由与这些政策行动无关的因素驱动的。

Second, the BVAR includes a large number of banking-sector endogenous variables: the amount and cost of credit, bank loan default rates, bank income statement variables, bank regulatory capital, and bank stock prices. This rich characterization of the banking sector is crucial for isolating the impact of regulatory and supervisory interventions on bank capital from the effects of other variables, as it allows us to consider several potential interactions between developments in the real economy and in financial and credit markets. Indeed, this strength translates into a fairly good out-of-sample forecasting performance of the model. In order to validate our choice of variables included in the VAR and to show that bank capital is adequately modeled, we discuss and report formal econometric evidence based on block exogeneity tests and forecasting assessment in Section 4.4 and in the Supporting Information Appendix.
其次，BVAR 包括大量银行部门内生变量：信贷的数量和成本、银行贷款违约率、银行收入报表变量、银行监管资本和银行股价。这种对银行部门的丰富描述对于将监管和监督干预对银行资本的影响与其他变量的影响分开至关重要，因为它使我们能够考虑实际经济与金融和信贷市场发展之间的几种潜在互动。实际上，这种优势转化为模型相当良好的样本外预测性能。为了验证我们在 VAR 中包含的变量选择，并证明银行资本得到了充分建模，我们在第 4.4 节和支持信息附录中讨论并报告基于区块外生性测试和预测评估的正式计量经济学证据。

Third, consistently with the literature on scenario analysis (Altavilla, Canova, and Ciccarelli 2020; Rostagno et al. 2019), in the second step of the analysis, we impose that (i) the path of monetary policy is the same in the policy and no-policy scenarios and that (ii) slow-moving variables (real GDP, HICP, and loans to NFCs and HHs) do not react on impact. These restrictions help to limit the combinations of shocks that can be used to satisfy the conditioning path (Rostagno et al. 2019) and are equivalent to assuming that policy/regulatory measures were the only driver of changes in the Tier 1 ratio on impact.
第三，与情景分析文献（Altavilla, Canova, 和 Ciccarelli 2020；Rostagno 等 2019）一致，在分析的第二步中，我们强加以下条件：(i) 政策和无政策情景下的货币政策路径相同，以及 (ii) 缓慢变化的变量（实际 GDP、HICP 和对非金融公司及家庭的贷款）不会立即反应。这些限制有助于限制可以用来满足条件路径的冲击组合（Rostagno 等 2019），并等同于假设政策/监管措施是影响一级资本比率变化的唯一驱动因素。

Our results show that both the 2011 EBA stress test and the 2014 CA and Introduction of the SSM increased the Tier 1 ratio of Italian banks by about 2.5 percentage points over a two-year horizon. The behavior of key macroeconomic variables suggests that these capital increases led to a tightening of credit supply conditions: On average over the two periods, the stock of loans to NFCs was lower by between

4.4 %

and

5.2 %

(at the end of the horizon, depending on the episode considered and whether we consider the counterfactual based on the unconditional forecast or the conditional one); the stock of loans to households declined significantly but only in the SSM/CA episode (by between

4.4 %

and 5.5%). Lending margins for both households and firms (i.e., the difference between lending rates and short-term interest rates) increased by between 80 and 150 basis points. As a result of the tighter credit conditions, GDP was significantly affected: In the EBA episode, the decline was concentrated in the first year after the policy action (with a peak effect of around

2 %

); in the SSM/CA episode, the largest impact is estimated at the end of the horizon, with an effect of between

3 %

and

4 %

(depending on the counterfactual considered). Finally, the reaction of consumer prices was different in these episodes, with a decrease in the EBA episode and an increase in the SSM/CA episode.
我们的结果表明，2011 年 EBA 压力测试和 2014 年 CA 及 SSM 的引入使意大利银行的一级资本比率在两年内提高了约 2.5 个百分点。关键宏观经济变量的行为表明，这些资本增加导致了信贷供应条件的收紧：在这两个时期的平均水平上，对非金融公司的贷款存量减少了

4.4 %

到

5.2 %

（在考虑的情境和我们是否考虑基于无条件预测或有条件预测的反事实的情况下，在时间范围结束时）；对家庭的贷款存量显著下降，但仅在 SSM/CA 情境中（减少了

4.4 %

到 5.5%）。家庭和企业的贷款利差（即贷款利率与短期利率之间的差额）增加了 80 到 150 个基点。由于信贷条件收紧，GDP 受到了显著影响：在 EBA 事件中，下降集中在政策行动后的第一年（峰值效应约为

2 %

）；在 SSM/CA 事件中，最大影响预计在时间范围的末尾，效应在

3 %

和

4 %

之间（取决于考虑的反事实）。最后，消费者价格在这些事件中的反应不同，EBA 事件中下降而 SSM/CA 事件中上升。

In contrast, the increase in the Tier 1 ratio terminating at the end of the two-year window following the Basel III reform is correctly predicted by our BVAR, implying that bank capital increases were consistent with macroeconomic and banking developments and signaling the good fitting properties of the model. Consistently, we find no effect on credit or macroeconomic variables during this episode.
与此相反，巴塞尔协议 III 改革后两年窗口结束时一级资本比率的增加被我们的 BVAR 正确预测，这意味着银行资本的增加与宏观经济和银行业的发展是一致的，并且表明模型的良好拟合特性。我们一致发现，在这一时期对信贷或宏观经济变量没有影响。

Our estimated effects on GDP are consistent with the results in the literature on the impact of bank capital shocks. Looking at some of the closest empirical papers to ours (Meeks 2017; Mésonnier and Stevanovic 2017; Mésonnier and Monks 2015; Kanngiesser et al. 2020) and normalizing the size of their increases in the Tier 1 ratio to the values obtained in this paper (2.5pp), the estimated impact on GDP is between

1 %

and

3 %

; the decline in loans to NFCs is larger than the average effect in our simulation in three of the four papers mentioned; the increase in credit spreads (considering both households and firms) is between 30 and 150 bps . The effects on loans and spreads are also similar in magnitude to those obtained in different settings by DelGiovane, Nobili, and Signoretti (2017) and Bassett, Lee, and Spiller (2015). Moreover, it should be kept in mind that our results are specific to episodes in which the narrative suggests that the size and impact of the policy actions are likely to have been greatest; this justifies potentially larger effects than those found in structural analysis, whose results typically represent averages across episodes.
我们对 GDP 的估计影响与文献中关于银行资本冲击影响的结果一致。查看一些与我们研究最接近的实证论文（Meeks 2017；Mésonnier 和 Stevanovic 2017；Mésonnier 和 Monks 2015；Kanngiesser 等人 2020），并将它们的一级资本比率增加的规模标准化为本文获得的值（2.5 个百分点），对 GDP 的估计影响在

1 %

和

3 %

之间；对非金融公司（NFCs）贷款的下降在提到的四篇论文中有三篇的平均效果更大；信贷利差的增加（考虑家庭和企业）在 30 到 150 个基点之间。贷款和利差的影响在不同环境下与 DelGiovane、Nobili 和 Signoretti（2017）以及 Bassett、Lee 和 Spiller（2015）获得的结果在数量上也相似。此外，应该记住，我们的结果特定于叙述表明政策行动的规模和影响可能最大的一些事件；这就解释了可能比结构分析中发现的效果更大的原因，后者的结果通常代表跨事件的平均值。

A potentially important concern with our empirical strategy is that the scenario analysis may not correctly attribute the increases in the Tier 1 ratio observed in our episodes to regulatory/ supervisory interventions. As noted above, there are reasons to believe this is the case: (i) the richness of the model and the goodness of fit, (ii) the narrative, and (iii) the monetary policy and GDP invariance under the actual and counterfactual path.
我们实证策略中一个潜在的重要问题是，情景分析可能未能正确归因于我们案例中观察到的一级资本充足率的增加与监管/监督干预之间的关系。如上所述，有理由相信情况确实如此：(i) 模型的丰富性和拟合优度，(ii) 叙述，和 (iii) 在实际和反事实路径下货币政策和 GDP 的不变性。

To address this concern, we conduct a number of additional analyses and robustness checks, all of which confirm that the results obtained in the scenario analysis are correctly attributed to the effect of increased regulatory/supervisory pressure on bank capital. First, we discuss complementary evidence from the ECB’s Bank Lending Survey (BLS), which since 2011 has included an ad hoc question asking banks whether new or expected regulatory and/or supervisory actions have had an impact on their capital position, distinguishing between the impact on equity and RWAs. This evidence strongly supports the notion that the increases in banks’ capital ratios during the EBA and SSM/CA episodes were driven by regulatory/supervisory actions. Second, we show that the actual developments of the main variables of interest fall within the credible interval of their (conditional) forecasts in the policy scenario, confirming that no other major events occurred in the windows and strengthening our case that the dynamics of the Tier 1 ratio can be attributed to the regulatory/supervisory actions. Third, we replicate the baseline simulations by adding the US Tier 1 ratio to the VAR. This variable captures a global banking factor that helps to isolate the evolution of the Italian Tier 1 ratio from developments unrelated to the “European” policies; the results confirm all the findings of the baseline model. Fourth, we perform a “placebo” test to verify that our model does not systematically deviate from the target variable (the Tier 1 ratio). Finally, we consider separately the numerator (Tier 1 capital) and the denominator (RWAs) of the Tier 1 ratio. Our results are confirmed. Moreover, we show that for the EBA episode, most of the effects were due to a decrease in RWAs, while for the SSM/CA episode, the effects were due to both an increase in Tier 1 capital and a decrease in RWAs.
为了解决这一问题，我们进行了多项额外分析和稳健性检验，所有结果均确认情景分析中获得的结果确实归因于监管/监督压力增加对银行资本的影响。首先，我们讨论来自欧洲央行银行信贷调查（BLS）的补充证据，自 2011 年以来，该调查包括一个临时问题，询问银行新或预期的监管和/或监督措施是否对其资本状况产生了影响，并区分对股本和风险加权资产（RWA）的影响。这一证据强烈支持在 EBA 和 SSM/CA 事件期间，银行资本比率的增加是由监管/监督措施驱动的观点。其次，我们展示了主要关注变量的实际发展落在政策情景下其（条件）预测的可信区间内，确认在这些时间窗口内没有发生其他重大事件，并加强了我们的论点，即一级资本比率的动态可以归因于监管/监督措施。第三，我们通过将美国一级资本比率添加到 VAR 中来复制基线模拟。该变量捕捉了一个全球银行因素，帮助将意大利的一级资本充足率的演变与与“欧洲”政策无关的发展隔离开来；结果确认了基线模型的所有发现。第四，我们进行了一项“安慰剂”测试，以验证我们的模型是否没有系统性地偏离目标变量（一级资本充足率）。最后，我们分别考虑一级资本充足率的分子（一级资本）和分母（风险加权资产）。我们的结果得到了确认。此外，我们显示，对于 EBA 事件，大多数影响是由于风险加权资产的减少，而对于 SSM/CA 事件，影响则是由于一级资本的增加和风险加权资产的减少。

Nevertheless, a crucial check is that the results of our scenario analysis in the EBA and SSM/CA episodes are consistent with those that would be obtained-using our sample and for the specific windows (i.e., not just compared to other papers and other samples)—via a structural VAR (SVAR) analysis, as we show in a companion paper (Conti, Nobili, and Signoretti 2023). Indeed, in Conti, Nobili, and Signoretti (2023), we estimate a SVAR model in which exogenous shocks to bank capital requirements are identified using narrative sign restrictions (Antolín-Díaz and Rubio-Ramírez 2018), a methodology that supports our approach as it exploits external information about particularly relevant events.

^{12}

The identification of bank capital requirement shocks using narrative sign restrictions in a SVAR model confirms the contractionary effects on lending supply and economic activity during the EBA and SSM forecast windows.
然而，一个关键的检查是，我们在 EBA 和 SSM/CA 事件中的情景分析结果与使用我们的样本和特定窗口（即，不仅仅是与其他论文和其他样本进行比较）通过结构向量自回归（SVAR）分析所获得的结果是一致的，正如我们在一篇配套论文中所展示的（Conti, Nobili, and Signoretti 2023）。实际上，在 Conti, Nobili, and Signoretti（2023）中，我们估计了一个 SVAR 模型，其中对银行资本要求的外生冲击是通过叙述性符号限制来识别的（Antolín-Díaz and Rubio-Ramírez 2018），这种方法支持了我们的方法，因为它利用了关于特别相关事件的外部信息。

^{12}

在 SVAR 模型中使用叙述性符号限制识别银行资本要求冲击确认了在 EBA 和 SSM 预测窗口期间对贷款供应和经济活动的收缩效应。

The rest of the paper is organized as follows. Section 2 reviews some contributions related to our work. Section 3 describes the evolution of the Tier 1 ratio in Italy. Section 4 outlines the empirical framework. Section 5 explains the procedure used to recover the size of the regulatory/supervisory actions. Section 6 discusses the impact on the amount and cost of lending and on the real economy. Section 7 shows the results obtained by allowing for time variation in coefficients and volatility. In Section 8, we provide several additional analyses and robustness checks. Finally, in Section 9, we conclude the study. A supporting information appendix is also available.
本文的其余部分组织如下。第二节回顾了与我们工作相关的一些贡献。第三节描述了意大利一级资本比率的演变。第四节概述了实证框架。第五节解释了用于恢复监管/监督行动规模的程序。第六节讨论了对贷款金额和成本以及对实体经济的影响。第七节展示了通过允许系数和波动性随时间变化而获得的结果。在第八节中，我们提供了几项额外分析和稳健性检验。最后，在第九节中，我们总结了研究。还提供了支持信息附录。

This paper connects to several strands of literature. First, it relates to the empirical works measuring the effect of regulatory/supervisory shocks based on direct observation of bankspecific capital requirements (Meeks 2017; Aiyar, Calomiris, and Wieladek 2016; De Jonghe, Dewachter, and Ongena 2020) and/or studies using exogenous bank-level losses and exploiting event studies (Jiménez et al. 2017; Mésonnier and Monks 2015; Gropp et al. 2019). In this regard, our approach provides an alternative method to construct a proxy of the impact of raising bank capital requirements when dealing with the scarcity or unobservability of confidential microdata on capital requirements. Compared to this literature, we allow for two-way feedback effects between bank capital requirements, macroeconomic and financial conditions, and other bank characteristics that cannot be accounted for in single-equation regressions. As for the interpretation of the results, our methodology is somewhat more general, as our estimation of the size of policy actions is not limited to exogenous variations in specific capital requirements but also includes the effect of a broad range of regulatory measures and supervisory pressures over a longer sample period.
本文与几条文献线索相关。首先，它与基于对银行特定资本要求的直接观察来衡量监管/监督冲击影响的实证研究相关（Meeks 2017；Aiyar, Calomiris, 和 Wieladek 2016；De Jonghe, Dewachter, 和 Ongena 2020），和/或使用外生银行级别损失并利用事件研究的研究（Jiménez et al. 2017；Mésonnier 和 Monks 2015；Gropp et al. 2019）。在这方面，我们的方法提供了一种替代方法，以构建提高银行资本要求影响的代理，当处理关于资本要求的机密微观数据的稀缺或不可观察性时。与这些文献相比，我们允许银行资本要求、宏观经济和金融条件以及其他银行特征之间的双向反馈效应，这在单方程回归中无法考虑。至于结果的解释，我们的方法论相对更为一般，因为我们对政策行动规模的估计不仅限于特定资本要求的外生变化，还包括在更长的样本期内广泛的监管措施和监督压力的影响。

Secondly, our estimation of the counterfactual series of capital ratio is conceptually similar to the notion of “economic capital”, that is, a pre-specified time-varying level of capitalization, consistent with business cycle and financial conditions, that banks target when choosing their actual level of capital Mésonnier and Stevanovic (2017); DeNicolò (2015); Berrospide and Edge (2010); Hancock and Wilcox (1994). This literature also needs confidential bank-level data with the obvious advantage of allowing for heterogeneity across banks and potentially controlling for all aggregate shocks when recovering the bank capital shocks.

^{13}

An important difference with our study is that in these papers, a positive (higher) difference between actual and “economic” capital is interpreted as reflecting banks’ ability to maintain (increase) a capital buffer beyond their target level. This interpretation relies crucially on the assumption (either implicitly or explicitly acknowledged) that in the period over which those models are estimated, the regulatory constraint on bank leverage is slack. Our approach, instead, does not require to make assumptions about regulatory/supervisory constraints which, by construction (since these are captured in our bank capital dynamic paths), may vary over time and affect all the variables in the model.
其次，我们对资本比率反事实系列的估计在概念上类似于“经济资本”的概念，即一个预先指定的、随时间变化的资本化水平，与商业周期和金融条件一致，这是银行在选择其实际资本水平时所针对的 Mésonnier 和 Stevanovic (2017); DeNicolò (2015); Berrospide 和 Edge (2010); Hancock 和 Wilcox (1994)。这方面的文献还需要机密的银行级数据，显然的优势在于允许银行之间的异质性，并在恢复银行资本冲击时可能控制所有的总量冲击。

^{13}

我们研究的一个重要区别在于，在这些论文中，实际资本与“经济”资本之间的正（更高）差异被解释为反映银行维持（增加）超出其目标水平的资本缓冲的能力。这种解释在很大程度上依赖于假设（无论是隐含还是明确承认）在这些模型估计的期间，银行杠杆的监管约束是宽松的。我们的方法则不需要对监管/监督约束做出假设，因为这些约束（由于它们被纳入我们的银行资本动态路径中）可能随时间变化并影响模型中的所有变量。

Third, our work is connected to papers that assess the macroeconomic effects of bank capital shocks using different methodologies, namely, DSGE models (Basel Committee on Banking Supervision 2010, 2015; Angelini et al. 2011; Mendicino et al. 2020) and VARs identified with zero and/or sign exclusion restrictions (e.g., Kanngiesser et al. 2020; Noss and Toffano 2016; Meeks 2017). We contribute to this field of research mainly by offering a different strategy to recover the size of regulatory/ supervisory interventions.

^{14}

In addition, having a significantly larger set of endogenous variables underlying the conditional forecast allows us to improve with respect to the quantitative evaluation of bank capital shocks, while models including a banking sector are typically limited to adding credit volumes and lending rates in small-scale VARs (Prieto, Eickmeier, and Marcellino 2016; Gambetti and Musso 2017). Finally, we address the issue of time variation in the estimated relationships. To the best of our knowledge, this is the first paper using time-varying VAR models to investigate the effects of regulatory tightening.
第三，我们的工作与评估银行资本冲击的宏观经济影响的论文相关，这些论文使用不同的方法论，即 DSGE 模型（巴塞尔银行监管委员会 2010 年，2015 年；Angelini 等 2011 年；Mendicino 等 2020 年）和使用零和/或符号排除限制识别的 VAR（例如，Kanngiesser 等 2020 年；Noss 和 Toffano 2016 年；Meeks 2017 年）。我们主要通过提供一种不同的策略来恢复监管/监督干预的规模，为这一研究领域做出贡献。此外，拥有显著更大的内生变量集以支持条件预测，使我们在银行资本冲击的定量评估方面有所改善，而包括银行部门的模型通常仅限于在小规模 VAR 中添加信贷量和贷款利率（Prieto, Eickmeier, 和 Marcellino 2016 年；Gambetti 和 Musso 2017 年）。最后，我们解决了估计关系中的时间变化问题。据我们所知，这是第一篇使用时间变化 VAR 模型来研究监管收紧影响的论文。

The paper also relates to the recently developed medium-scale Bayesian VAR models that are suitable to address the curse of dimensionality and whose typical application is counterfactual simulations aimed at detecting misalignments and irregularities in the observable developments of macroeconomic variables (Giannone et al. 2012; Giannone, Lenza, and Reichlin 2019; DeMol, Giannone, and Reichlin 2008; Aastveit et al. 2017; Bobeica and Jarociński 2019; Caruso, Reichlin, and Ricco 2019). Our approach has much in common with the recent class of models studying the monetary transmission mechanism and credit shocks with medium- and large-scale VARs (von Borstel, Eickmeier, and Krippner 2016; Boivin, Giannoni, and Stevanovic 2020). However, as discussed below, unlike these papers, we do not consider impulse responses derived by a structural model but instead use a conditional forecasting approach, akin to scenario analysis such as in Lenza, Pill, and Reichlin (2010); Giannone et al. (2012); Altavilla, Giannone, and Lenza (2016); Altavilla, Canova, and Ciccarelli (2020).

Finally, our analysis builds upon the specification of single-equation models typically used in central banks for the analysis of credit market developments (Albertazzi et al. 2014; Bofondi and Ropele 2011); an obvious advantage of a multivariate approach is the ability to model a large number of endogenous variables in a unified framework.
本文还涉及最近开发的中等规模贝叶斯 VAR 模型，这些模型适合解决维度诅咒，其典型应用是反事实模拟，旨在检测宏观经济变量可观察发展中的错位和不规则性（Giannone et al. 2012; Giannone, Lenza, and Reichlin 2019; DeMol, Giannone, and Reichlin 2008; Aastveit et al. 2017; Bobeica and Jarociński 2019; Caruso, Reichlin, and Ricco 2019）。我们的方法与最近一类研究货币传导机制和信贷冲击的中大规模 VAR 模型有很多共同之处（von Borstel, Eickmeier, and Krippner 2016; Boivin, Giannoni, and Stevanovic 2020）。然而，如下所述，与这些论文不同，我们不考虑由结构模型推导的脉冲响应，而是使用条件预测方法，类似于情景分析，如在 Lenza, Pill, and Reichlin (2010); Giannone et al. (2012); Altavilla, Giannone, and Lenza (2016); Altavilla, Canova, and Ciccarelli (2020)中所述。

最后，我们的分析基于中央银行通常用于分析信贷市场发展的单方程模型的规范（Albertazzi et al. 2014; Bofondi and Ropele 2011）；多元方法的一个明显优势是能够在统一框架中建模大量内生变量。

3 | Institutional Background and the Evolution of Tier 1 Capital Ratio
3 | 机构背景与一级资本充足率的演变

A number of important milestones characterize the evolution of capital regulation and supervision of Italian banks in the sample covered by this study. First, the adoption of legislation introducing the Basel 2 standards in 2006 modified the criteria for calculating RWAs (most notably, introducing internal-rating based methodologies) and introduced the possibility for supervisory authorities to impose-beyond the minimum capital requirement that applies to all banks (so-called “Pillar 1”)—additional requirements (“Pillar 2”) based on an institution-specific assessment of the risks that are not adequately covered by Pillar 1 requirements. Second, the adoption of the Basel 3 package, following the Global Financial crisis determined an increase in (Pillar 1) required capital, introduced a minimum requirement for common equity, additional macroprudential capital buffers, a (non-risk-weighted) leverage ratio, and revisions to the calculation of risk weights; moreover, the reform introduced supplemental (Pillar 2) requirements addressing, among other things, firm governance and risk management and interest rate risk. Third, since November 2014, banking supervision of the largest Italian banks passed from the sole responsibility of the Bank of Italy to being competence of the SSM.

^{16}

本研究所涵盖的意大利银行资本监管和监督的演变中有几个重要的里程碑。首先，2006 年引入巴塞尔 2 标准的立法修改了计算风险加权资产（RWAs）的标准（最显著的是引入了基于内部评级的方法），并引入了监管机构在适用于所有银行的最低资本要求（所谓的“第一支柱”）之外，基于对未被第一支柱要求充分覆盖的风险的机构特定评估，施加额外要求（“第二支柱”）的可能性。其次，巴塞尔 3 方案的采用，紧随全球金融危机之后，导致了（第一支柱）所需资本的增加，引入了普通股的最低要求、额外的宏观审慎资本缓冲、（非风险加权）杠杆比率，以及对风险权重计算的修订；此外，改革引入了补充的（第二支柱）要求，涉及公司治理、风险管理和利率风险等方面。第三，自 2014 年 11 月起，意大利最大的银行的银行监管从意大利银行的单独责任转移至 SSM 的职能。

^{16}

Figure 1 plots the evolution of the aggregate Tier 1 capital ratio (upper panel) for Italian banks since 1994. The aggregate ratio is obtained as a weighted average of all the banking groups and individual institutions resident in Italy, on a consolidated basis.

^{17}

Three distinct phases can be observed in the evolution of the Tier 1 ratio in our sample: (i) the second half of the 1990s, when the Tier 1 ratio declined from around

10 %

to below

8 %

; (ii) the 2000s until the financial crisis, when the ratio hovered in a narrow range (between

7.5 %

and

9 %

); and (iii) the crisis and postcrisis period, when the ratio showed a sharp and relatively steady increase, reaching almost

13 %

at the end of the sample (2015:Q4).
图 1 绘制了自 1994 年以来意大利银行的整体一级资本充足率（上图）的演变。整体比率是通过对所有在意大利的银行集团和个别机构的加权平均得出的，基于合并数据。

^{17}

在我们的样本中，可以观察到一级资本充足率演变的三个不同阶段：（i）1990 年代后半期，一级资本充足率从大约

10 %

下降到低于

8 %

；（ii）2000 年代直到金融危机，比例在一个狭窄的范围内徘徊（在

7.5 %

和

9 %

之间）；以及（iii）危机及后危机时期，比例显示出急剧且相对稳定的增长，到样本结束时（2015:Q4）几乎达到了

13 %

。

Our analysis focuses on the most recent period. Since 2009, the increase in capitalization reflected both the steady decline in RWAs and the increase in equity capital, which was particularly strong up to 2012 (Figure 1, bottom panel).

^{18}

During that period, the evolution of the Tier 1 ratio was influenced by a number of important regulatory innovations and supervisory initiatives. In particular, we analyze three windows of 2 years during which regulatory and/or supervisory initiatives have raised pressure on banks to increase capitalization, and at the same time, large and persistent increases in the Tier 1 ratio were observed.
我们的分析集中在最近的时期。自 2009 年以来，资本化的增加反映了风险加权资产（RWAs）的稳步下降和股本的增加，特别是在 2012 年之前的增长非常强劲（图 1，底部面板）。

^{18}

在此期间，一级资本比率的演变受到了一些重要监管创新和监督举措的影响。特别是，我们分析了三个为期 2 年的窗口，在这些窗口期间，监管和/或监督举措对银行增加资本化施加了压力，同时观察到了一级资本比率的大幅和持续增长。

The first window starts in 2009:Q2 and includes the period following the start of the discussion on the Basel III regulatory reform. In the aftermath of the global financial crisis, international cooperation aimed at strengthening financial regulation and supervision intensified. Since 2008, preparatory work involved the Group of Twenty (G20), the Financial Stability Board (FSB), and the European Union and led to a number of recommendations that started to put pressures on the capitalization of the banking system. In November 2008, the Basel Committee on Banking Supervision (BCBS) approved an action plan whose primary objective was to “strengthen capital buffers and help contain leverage in the banking system […].” $^{19}$ . At the meetings in April and November 2009, the leaders of the G20 countries committed to completing a global reform of prudential regulation (Banca d’Italia, 2010). In December 2009, the Basel Committee on Banking Supervision published a consultation document with concrete proposals for capital and liquidity regulatory reforms (Basel Committee on Banking Supervision 2009) (BCBS, 2009). The final text of the Basel III reform was approved at the end of 2010 (BCBS, 2010b). $^{20}$
第一个窗口始于 2009 年第二季度，包括巴塞尔 III 监管改革讨论开始后的时期。在全球金融危机之后，旨在加强金融监管和监督的国际合作加剧。自 2008 年以来，准备工作涉及二十国集团（G20）、金融稳定委员会（FSB）和欧盟，并提出了一系列建议，开始对银行系统的资本化施加压力。2008 年 11 月，巴塞尔银行监管委员会（BCBS）批准了一项行动计划，其主要目标是“加强资本缓冲并帮助控制银行系统的杠杆率……”。 $^{19}$ 。在 2009 年 4 月和 11 月的会议上，G20 国家的领导人承诺完成全球审慎监管改革（意大利银行，2010）。2009 年 12 月，巴塞尔银行监管委员会发布了一份咨询文件，提出了关于资本和流动性监管改革的具体建议（巴塞尔银行监管委员会 2009）（BCBS，2009）。巴塞尔 III 改革的最终文本于 2010 年底获得批准（BCBS，2010b）。 $^{20}$
The second window starts in 2011:Q1 and covers (i) the 2011 stress test run by the European Banking Authority (EBA), launched in January 2011, and (ii) the EBA one-off Capital exercise, which was announced in October of the same year following a decision by the European Council. The stress test was based on banks’ balance sheets as of December 2010, and the results were published in July 2011. The aim of the stress test was to assess the resilience of the banks involved in the exercise against an adverse but plausible scenario. This episode determined an increase in capital of participating banks in three ways. First, the EBA allowed specific capital increases in the first 4 months of 2011 (i.e., before the completion of the stress test exercise, in July) to be considered in the results; banks were therefore incentivized to strengthen their capital positions ahead of the stress test. Second, the EBA has issued a formal recommendation stating that national supervisory authorities should require banks whose Core Tier 1 ratio falls below the $5 %$ threshold to promptly remedy their capital shortfall. Third, the EBA also recommended that national supervisory authorities request all banks whose Cet 1 ratio is above but close to $5 %$ , and which have sizeable exposures to sovereigns under stress, to take specific steps to strengthen their capital position, including restrictions on dividends, deleveraging, issuance of fresh capital or conversion of lower-quality instruments into Core Tier 1 capital. In anticipation of the stress test results, between January and April 2011, about 50 billion capital was raised on a net basis by the 90 EU banks participating $^{21}$ ;about 11 billion was the amount raised by the five largest Italian banks participating (which amounts to about $1 %$ of RWAs at the end of 2010. $^{22}$ In addition, for four out of the five Italian banks participating to the exercise, the EBA identified a total capital shortfall of 15.4 billion, which banks were prescribed to cover by the end of June 2012. $^{23}$ Overall, the Tier 1 ratio in the last quarter of 2011 and in 2012 increased by 1.1 percentage points.
第二个窗口始于 2011 年第一季度，涵盖了(i) 2011 年由欧洲银行管理局（EBA）进行的压力测试，该测试于 2011 年 1 月启动，以及(ii) EBA 的一次性资本评估，该评估是在同年 10 月根据欧洲理事会的决定宣布的。压力测试基于截至 2010 年 12 月的银行资产负债表，结果于 2011 年 7 月公布。压力测试的目的是评估参与测试的银行在不利但合理的情景下的韧性。这一事件通过三种方式决定了参与银行资本的增加。首先，EBA 允许在 2011 年头四个月（即在压力测试完成之前的 7 月）进行特定的资本增加，以便在结果中考虑；因此，银行被激励在压力测试之前加强其资本状况。其次，EBA 发布了一项正式建议，指出国家监管机构应要求核心一级资本比率低于 $5 %$ 阈值的银行迅速弥补其资本不足。第三，EBA 还建议国家监管机构要求所有核心一级资本比率高于但接近 $5 %$ 的银行，以及对处于压力下的主权债务有大量敞口的银行，采取具体措施加强其资本状况，包括限制分红、去杠杆、发行新资本或将低质量工具转换为核心一级资本。为了预期压力测试结果，在 2011 年 1 月至 4 月期间，90 家参与的欧盟银行净筹集了约 500 亿资本 $^{21}$ ；五家参与的意大利最大银行筹集的金额约为 110 亿（这相当于 2010 年底 RWAs 的约 $1 %$ ）。 $^{22}$ 此外，对于参与该项工作的五家意大利银行中的四家，EBA 识别出总资本缺口为 154 亿，要求银行在 2012 年 6 月底之前弥补。 $^{23}$ 总体而言，2011 年最后一个季度和 2012 年的一级资本比率提高了 1.1 个百分点。
The third time window starts in 2014:Q1 and covers the period of the ECB’s CA and the first months of operation of the SSM. The CA was announced in October 2013 and was run in the subsequent 12 months; it consisted in a thorough “health check” of the European banks (the Asset Quality Review, AQR ) and a stress test, which were conducted ahead of the start of SSM operation at the beginning of November 2014. In the CA, additional capital needs (shortfalls) were identified as a result of both the AQR and following the result of the stress test, in the case in which capital in the adverse scenario fell below the minimum threshold (of ). Similar to the 2011 EBA exercise, capital increases that occurred before the completion of the CA (in the second half of 2013 and in 2014) were de facto allowed to cover capital shortfalls; thus, banks were incentivized to strengthen their capital ahead of the CA. In anticipation of the CA results, Italian banks undertook significant measures that strengthened their capitalization by about 15 billion; also taking into account these measures, the results envisaged further aggregated capital needs of about 3 billion euro. In 2015, which coincided with the first year of operation of the SSM, additional capital increases were recorded. Overall, in 2014 and 2015 aggregate equity capital of Italian banks increased by about 20 billion.
第三个时间窗口始于 2014 年第一季度，涵盖了欧洲央行的 CA 和 SSM 运营的前几个月。CA 于 2013 年 10 月宣布，并在随后的 12 个月内进行；它包括对欧洲银行的全面“健康检查”（资产质量审查，AQR）和压力测试，这些测试是在 2014 年 11 月初 SSM 运营开始之前进行的。在 CA 中，额外的资本需求（短缺）是由于 AQR 和压力测试结果而确定的，在不利情景下资本低于最低阈值（）的情况下。与 2011 年 EBA 的演习类似，在 CA 完成之前（2013 年下半年和 2014 年）发生的资本增加实际上被允许用于弥补资本短缺；因此，银行被激励在 CA 之前加强其资本。为了预期 CA 结果，意大利银行采取了显著措施，使其资本化增强了约 150 亿欧元；考虑到这些措施，结果预计进一步的资本需求约为 30 亿欧元。 2015 年，恰逢 SSM 的第一年运营，记录了额外的资本增加。总体而言，2014 年和 2015 年意大利银行的总股本增加了约 200 亿。

Tier 1 capital ratio
一级资本充足率

Notes: Top panel: Bank Tier 1 ratio, percentage points. Bottom panel: the black lines represent bank Tier 1 capital (left) and the RiskWeighted Assets (right); the cyan shaded areas denote the first quarter of each considered episode (Basel III; EBA Stress Test; Launch of the SSM and the ECB Comprehensive Assessment). Source: Bank of Italy credit register. Quarterly data for the period 1993-2015.
注释：顶部面板：银行一级资本比率，百分比点。底部面板：黑线代表银行一级资本（左）和风险加权资产（右）；青色阴影区域表示每个考虑的事件的第一季度（巴塞尔协议 III；欧洲银行管理局压力测试；单一监管机制的启动和欧洲中央银行全面评估）。来源：意大利银行信用登记。1993-2015 年期间的季度数据。

FIGURE 1 | The evolution of bank capital in Italy.
图 1 | 意大利银行资本的演变。

4 | Empirical Framework
4 | 实证框架

4.1 | Fixed-Coefficients Bayesian VAR Model
4.1 | 固定系数贝叶斯 VAR 模型

To address our research question, we use a BVAR model, which provides us with a flexible tool to deal with the interlinkages between macroeconomic, financial, and banking variables without imposing too much structure on the data.

^{26}

Our reference model is given by
为了回答我们的研究问题，我们使用 BVAR 模型，这为我们提供了一个灵活的工具，以处理宏观经济、金融和银行变量之间的相互联系，而不对数据施加过多的结构。

^{26}

我们的参考模型如下

Y_{t} = A_{0} + A_{1} Y_{t - 1} + A_{2} Y_{t - 2} + \dots + A_{p} Y_{t - p} + ε_{t}, ε_{t} \sim N (0, Σ)

or in terms of polynomial matrix form,

Y_{t} = B (L) Y_{t - 1} + ε_{t}

, and equivalently, in even more compact form,
或用多项式矩阵形式表示，

Y_{t} = B (L) Y_{t - 1} + ε_{t}

，并且等效地，以更紧凑的形式，

Y_{t} = X_{t}^{'} B + ε_{t}

where

Y_{t}

is a

m \times 1

vector of endogenous variables,

A_{0}, \dots, A_{p}

are

m \times m

matrix of coefficients, and

ε_{t}

is a vector of residuals, which are assumed to be normally distributed with zero mean and variance-covariance matrix

Σ

and where

X_{t}

contains the constant and the lags of the endogenous variables, whereas

B

contains the matrices

A_{0}, \dots, A_{p}

.
其中

Y_{t}

是一个

m \times 1

内生变量向量，

A_{0}, \dots, A_{p}

是

m \times m

系数矩阵，

ε_{t}

是一个残差向量，假设其服从均值为零、方差-协方差矩阵为

Σ

的正态分布，且

X_{t}

包含常数项和内生变量的滞后项，而

B

包含矩阵

A_{0}, \dots, A_{p}

。

Since our aim is to exploit a large amount of information to capture the correlations between the main macroeconomic variables and the banking sector, we choose a Bayesian framework, which is particularly useful when dealing with large systems of variables (Banbura, Giannone, and Reichlin 2010; Giannone et al. 2012; Banbura, Giannone, and Lenza 2015). This class of models, indeed, allows to attenuate overfitting problems, performs very well in terms of out-of-sample forecasts, and provides reliable impulse responses of the main macroeconomic variables in the euro area to structural shocks (see Giannone et al. 2014; Banbura, Giannone, and Lenza 2015). We keep the prior settings as standard in the literature, by using the so-called sum of coefficients prior (Sims and Zha 1998; Banbura, Giannone, and Reichlin 2010). The details of the elicitation of priors are reported in the Supporting Information Appendix A, mainly relying on the papers by Giannone, Lenza, and Primiceri (2015) and Clark and McCracken (2014).
由于我们的目标是利用大量信息捕捉主要宏观经济变量与银行部门之间的相关性，我们选择了贝叶斯框架，这在处理大量变量系统时特别有用（Banbura, Giannone, and Reichlin 2010; Giannone et al. 2012; Banbura, Giannone, and Lenza 2015）。这一类模型确实可以减轻过拟合问题，在样本外预测方面表现良好，并提供欧元区主要宏观经济变量对结构性冲击的可靠脉冲响应（见 Giannone et al. 2014; Banbura, Giannone, and Lenza 2015）。我们将先验设置保持为文献中的标准，通过使用所谓的系数和先验（Sims and Zha 1998; Banbura, Giannone, and Reichlin 2010）。先验的引导细节在支持信息附录 A 中报告，主要依赖于 Giannone, Lenza 和 Primiceri（2015）以及 Clark 和 McCracken（2014）的论文。

4.2 | Time-Varying Coefficients Bayesian VAR Model
4.2 | 时变系数贝叶斯 VAR 模型

When we switch to the time-varying coefficients case to address the possibility of changes in the propagation or in the volatility of the exogenous disturbances to bank capital, we are constrained by the relatively large dimension of our system of variables which prevents us from using the seminal approach by Cogley and Sargent (2005) and Primiceri (2005). Instead, we closely follow the steps by Aastveit et al. (2017), who deal with 13 variable models of the US economy by sticking to the methodology proposed by Koop and Korobilis (2013). Model (1) is now allowed to have both time-varying coefficients and volatility, and it is estimated by Kalman filter techniques, as it can be casted in state space form:
当我们转向时间变动系数的情况，以解决外生干扰对银行资本的传播或波动变化的可能性时，我们受到变量系统相对较大维度的限制，这使我们无法使用 Cogley 和 Sargent（2005）以及 Primiceri（2005）提出的开创性方法。相反，我们紧密遵循 Aastveit 等人（2017）的方法，他们通过坚持 Koop 和 Korobilis（2013）提出的方法，处理美国经济的 13 个变量模型。模型（1）现在允许具有时间变动系数和波动性，并通过卡尔曼滤波技术进行估计，因为它可以被表示为状态空间形式：

Y_{t} = A_{0, t} + A_{1, t} Y_{t - 1} + A_{2, t} Y_{t - 2} + \dots + A_{p, t} Y_{t - p} + ε_{t}, ε_{t} \sim N (0, Σ_{t})

Y_{t} = B_{t} (L) Y_{t - 1} + ε_{t}

, or, equivalently, in even more compact form
或

Y_{t} = B_{t} (L) Y_{t - 1} + ε_{t}

，或者，更简洁地说，

\begin{matrix} Y_{t} = X_{t}^{'} B_{t} + ε_{t} \\ B_{t} = B_{t - 1} + η_{t}, var (η_{t}) = Q_{t} \end{matrix}

Koop and Korobilis (2013) propose to approximate the Kalman filtering formula for the state variance

V_{t ∣ t - 1} = V_{t - 1 ∣ t - 1} + Q_{t}

with

V_{t ∣ t - 1} = \frac{1}{λ} V_{t - 1 ∣ t - 1}

, that is, by introducing a forgetting factor

0 < λ \leq 1

in order to eliminate the need for estimating or simulating the matrix

Q_{t}

, which is particularly cumbersome and intensive from a computational point of view. A similar approximation is then used to bypass the need for a posterior simulation algorithm for multivariate stochastic volatility in the measurement equation. Indeed, Koop and Korobilis (2013) use an exponentially weighted moving average (EWMA) to model volatility estimator for the measurement error covariance matrix:

{\hat{Σ}}_{t} = κ {\hat{Σ}}_{t - 1} + (1 - κ) {\hat{ε}}_{t} {\hat{ε}}_{t}^{'}

, where

{\hat{ε}}_{t} = Y_{t} - X_{t} B_{t ∣ t}

is obtained with the Kalman filter. Following the baseline settings of Koop and Korobilis (2013), we set the forgetting factor

λ

at 0.99 and the volatility weighting coefficient

κ

at 0.96.

^{27}

Koop 和 Korobilis（2013）建议用

V_{t ∣ t - 1} = \frac{1}{λ} V_{t - 1 ∣ t - 1}

来近似状态方差

V_{t ∣ t - 1} = V_{t - 1 ∣ t - 1} + Q_{t}

的卡尔曼滤波公式，即通过引入遗忘因子

0 < λ \leq 1

来消除估计或模拟矩阵

Q_{t}

的需要，这在计算上尤其繁琐和密集。然后使用类似的近似方法来绕过测量方程中多元随机波动的后验模拟算法的需要。实际上，Koop 和 Korobilis（2013）使用指数加权移动平均（EWMA）来建模测量误差协方差矩阵的波动性估计器：

{\hat{Σ}}_{t} = κ {\hat{Σ}}_{t - 1} + (1 - κ) {\hat{ε}}_{t} {\hat{ε}}_{t}^{'}

，其中

{\hat{ε}}_{t} = Y_{t} - X_{t} B_{t ∣ t}

是通过卡尔曼滤波获得的。根据 Koop 和 Korobilis（2013）的基线设置，我们将遗忘因子

λ

设定为 0.99，将波动性加权系数

κ

设定为 0.96。

^{27}

4.3 | The Data
4.3 | 数据

The specification of the VAR model is designed to capture the most relevant interrelations between the banking system and the macroeconomy. Our dataset includes quarterly information on the Italian banking sector going back as far as possible and is rich enough to capture four recessions: the one in the early 1990s, the one in the early 2000s, the one following the Global Financial Crisis of 2008-2009, and the one following the Sovereign Debt Crisis of 2011-2012.
VAR 模型的规范旨在捕捉银行系统与宏观经济之间最相关的相互关系。我们的数据集包括尽可能早的意大利银行业季度信息，并且足够丰富以捕捉四次衰退：1990 年代初的衰退、2000 年代初的衰退、2008-2009 年全球金融危机后的衰退，以及 2011-2012 年主权债务危机后的衰退。

The choice of the endogenous variables takes as a starting point the growing literature on the impact of credit shocks on the business cycle (see, e.g., Prieto, Eickmeier, and Marcellino 2016; Gambetti and Musso 2017). These works typically limit themselves to adding credit volumes and rates to the usual macroeconomic variables considered in the New Keynesian model (output, prices and the policy rate). In addition, our paper includes information on the long-term rate, loan default rates, the main items of banks’ profitability, bank stock prices, Tier 1 capital, and RWAs. Including a large number of banking variables allows for a rich representation of the interactions between financial intermediation and the business cycle (see below, Section 4.4). The baseline specification includes 16 variables: (i) four core macroeconomic and financial variables: Italian real GDP; harmonized consumer prices; a measure of the short-term interest rate (the 3-month Euribor rate), which captures conventional monetary policy, as well as the financial strains originated in the interbank market during the Global Financial Crisis; a measure of the long-term interest rate (the yield on the 10-year Italian government bond), which reflects developments in both the long-term risk-free interest rate-and is thus affected by unconventional monetary policy measures-as well as changes in risk premia (which were particularly relevant during the sovereign debt crisis

^{28}

; (iii) six variables related to the credit market: the cost and the volume of loans to both nonfinancial corporations (NFCs) and households (HHs) for house purchase

^{29}

; default rates of both NFCs and

{HHs}^{30}

; (iv) four variables related to the main items of bank income statement: net interest income; noninterest income; operational expenses; loan loss provisions

^{31}

; (v) the Italian bank stock market index

^{32}

; and (vi) the Tier 1 capital ratio.

^{33}

In addition to the four macroeconomic variables and the bank stock price index, respectively, downloaded from the ECB Statistical Data Warehouse and from Refinitiv, the other variables are taken from the Bank of Italy supervisory reports. In the baseline model, all the variables enter in loglevels, with the exception of interest rates, default rates, and the Tier 1 ratio, which are expressed in levels. Figure B1 in the Supporting Information Appendix provides a graphical representation of all the series used in the model.
内生变量的选择以关于信贷冲击对商业周期影响的日益增长的文献为起点（参见，例如，Prieto, Eickmeier, 和 Marcellino 2016；Gambetti 和 Musso 2017）。这些研究通常仅限于将信贷量和利率添加到新凯恩斯模型中考虑的通常宏观经济变量（产出、价格和政策利率）中。此外，我们的论文还包括有关长期利率、贷款违约率、银行盈利能力的主要项目、银行股价、一级资本和风险加权资产（RWAs）的信息。包括大量银行变量可以丰富金融中介与商业周期之间相互作用的表现（见下文，第 4.4 节）。基准规范包括 16 个变量：（i）四个核心宏观经济和金融变量：意大利实际 GDP；协调消费者价格；短期利率的一个衡量指标（3 个月 Euribor 利率），它捕捉了常规货币政策以及在全球金融危机期间源于银行间市场的金融压力；长期利率的一个衡量指标（10 年期意大利政府债券收益率），它反映了长期无风险利率的发展，因此受到非常规货币政策措施的影响，以及风险溢价的变化（在主权债务危机期间尤其相关

^{28}

；（iii）与信贷市场相关的六个变量：非金融企业（NFCs）和家庭（HHs）购房贷款的成本和数量

^{29}

；NFCs 和

{HHs}^{30}

的违约率；（iv）与银行收入报表主要项目相关的四个变量：净利息收入；非利息收入；运营费用；贷款损失准备金

^{31}

；（v）意大利银行股票市场指数

^{32}

；以及（vi）一级资本充足率。

^{33}

除了从欧洲中央银行统计数据仓库和 Refinitiv 下载的四个宏观经济变量和银行股票价格指数外，其他变量均来自意大利银行的监管报告。在基准模型中，所有变量以对数水平输入，利率、违约率和一级资本比率除外，这些以水平表示。支持信息附录中的图 B1 提供了模型中使用的所有系列的图形表示。

The estimation sample runs from 1993:Q1 to 2015:Q4. The choice of the period is constrained by the availability of high-quality information for some banking variables, namely, the capital ratio, default rates, and measures of bank profitability. To check the robustness of our results, we replicate the exercises in the paper by considering the first-differences of the variables expressed in stocks. Accordingly, we specify in the model a prior mean lower than the one on the own lag of the dependent variable, which is consistent with the evolution of stationary variables.

^{34}

All the main results are confirmed.
估计样本从 1993 年第一季度到 2015 年第四季度。选择该时期受到某些银行变量的高质量信息可用性的限制，即资本比率、违约率和银行盈利能力的衡量标准。为了检查我们结果的稳健性，我们通过考虑以存量表示的变量的一阶差分来复制论文中的实验。因此，我们在模型中指定一个低于因变量自身滞后值的先验均值，这与平稳变量的演变是一致的。

^{34}

所有主要结果均得到确认。

4.4 | How Are Data Selected?
4.4 | 数据是如何选择的？

The choice of the variables included in the model is based on both theoretical and empirical arguments and is driven by the need to properly take into account all the main determinants of the Tier 1 ratio. Indeed, adequately modeling of the drivers of bank capital is a prerequisite for the scenario analysis to provide meaningful results.
模型中包含的变量选择基于理论和实证论据，并受到正确考虑所有一级资本比率主要决定因素的需要驱动。实际上，充分建模银行资本的驱动因素是情景分析提供有意义结果的前提。

One crucial reference for the determinants of bank capital in the literature is the contributions that rely on the notion of “economic capital” (Mésonnier and Stevanovic 2017; DeNicolò 2015; Berrospide and Edge 2010; Hancock and Wilcox 1994, see the discussion in Section 2). These papers typically estimate a dynamic model bank of bank capital-toassets ratios, in which a target ratio is function of bank-specific characteristics and macrofinancial variables. Bank-level drivers of bank capital usually include a measure of bank size (i.e., total assets), a measure of bank profitability (e.g., ROA), a measure of asset risk (e.g., the share of charge-offs to total assets), and/or some measure of asset structure (e.g., the share of loans in total assets). Macrofinancial variables typically include (current or expected) GDP, a measure of the short-term interest rate, a measure of financial spread, stock prices, or stock price volatility.
文献中关于银行资本决定因素的一个重要参考是依赖于“经济资本”概念的贡献（Mésonnier 和 Stevanovic 2017；DeNicolò 2015；Berrospide 和 Edge 2010；Hancock 和 Wilcox 1994，见第 2 节的讨论）。这些论文通常估计一个动态模型银行的资本与资产比率，其中目标比率是银行特定特征和宏观金融变量的函数。银行资本的银行层面驱动因素通常包括银行规模的一个衡量指标（即总资产）、银行盈利能力的一个衡量指标（例如，ROA）、资产风险的一个衡量指标（例如，坏账占总资产的比例）和/或某种资产结构的衡量指标（例如，贷款在总资产中的比例）。宏观金融变量通常包括（当前或预期的）GDP、短期利率的一个衡量指标、金融利差的一个衡量指标、股票价格或股票价格波动性。

As described above, our endogenous variables include, but are not limited to, all the drivers of bank capital typically used in this literature. Regarding bank-specific drivers of capital, we use four different items of banks’ income statement to account for developments in bank profitability (net interest income, loan loss provisions, operating costs, and and other income); as a measure of asset riskiness, we separately include default rates on loans to both households and firms; developments in loans to households and firms account for the evolution of banks’ asset structure and, in particular, are strongly correlated with RWAs; in addition, we also include interest rates on loans to households and firms. In terms of the macro-financial variables, we include real GDP and inflation, which are standard determinants of the business cycle; the short-term interest rate, which accounts for conventional monetary policy; the 10-year government bond yield, which takes into account the large swings in financial conditions on the sovereign debt market between 2011 and 2012; and bank stock prices, which capture the conditions for new equity issuance, the potential impact of tensions in financial markets on bank capital, and the role of “market discipline” forcing banks to strengthen their capital positions at times of concern about their resilience.
如上所述，我们的内生变量包括但不限于本领域文献中通常使用的所有银行资本驱动因素。关于银行特定的资本驱动因素，我们使用银行收入报表中的四个不同项目来考虑银行盈利能力的发展（净利息收入、贷款损失准备、运营成本和其他收入）；作为资产风险的衡量标准，我们分别包括对家庭和企业贷款的违约率；对家庭和企业贷款的发展反映了银行资产结构的演变，特别是与风险加权资产（RWA）有很强的相关性；此外，我们还包括对家庭和企业贷款的利率。在宏观金融变量方面，我们包括实际 GDP 和通货膨胀，这些是商业周期的标准决定因素；短期利率，考虑到传统货币政策；10 年期国债收益率，考虑到 2011 年至 2012 年主权债务市场金融条件的大幅波动；以及银行股票价格，反映了新股发行的条件、金融市场紧张对银行资本的潜在影响，以及“市场纪律”在对银行韧性产生担忧时迫使银行加强资本状况的作用。

Econometric tests confirm that our choice of the variables included in the baseline specification is able to adequately model the behavior of bank capital over our sample. First, we show that the correlations between the Tier 1 ratio, on the one hand, and the main components of income (in our specification, netinterest income and other income) and RWAs (loans to NFC and HH ), on the other hand, are consistent with prior expectations on their relationship. To this end, we run the following OLS regressions:
计量经济学测试确认，我们在基准规范中选择的变量能够充分建模样本期间银行资本的行为。首先，我们展示了一级资本比率与收入的主要组成部分（在我们的规范中，净利息收入和其他收入）以及风险加权资产（对非金融公司和家庭的贷款）之间的相关性，与我们对其关系的先前预期是一致的。为此，我们进行以下 OLS 回归：

Δ k_{t} = α + ρ Δ k_{t - 1} + β (L) Δ x_{t - 1} + ε_{t}, ε_{t} \sim n.i.d. (0, σ_{ε}^{2})

where

k_{t}

is bank capital,

k_{t - 1}

its first lagged value, and

x_{t}

are net interest income, other income, loans to NFC, and loans to households. Regressions are estimated in the first differences to take into account possible nonstationary behavior (which in the VAR is addressed by means of the priors). We estimate one regression for each dependent variable and for each of the three episodes that we consider in the paper (meaning that we cut the estimation sample at the end of the time window of each of the three episodes). Additional details on the equation specifications are provided in the Supporting Information Appendix C. The results of the regressions, reported in Tables C1 and C2 in the Supporting Information Appendix C, show that the estimated correlations are consistent with prior expectations about the relationship between the main components of net income and RWAs and the Tier 1 ratio. In particular, we find positive coefficients for the net interest income and other income and negative ones for loans to households and firms. The coefficients are statistically significant for all variables and episodes, with the only exception of loans to NFCs in the Basel 3 episode.
其中

k_{t}

是银行资本，

k_{t - 1}

是其第一个滞后值，

x_{t}

是净利息收入、其他收入、对非金融公司（NFC）的贷款和对家庭的贷款。回归是在第一差分中估计的，以考虑可能的非平稳行为（在 VAR 中通过先验来解决）。我们为每个因变量和我们在论文中考虑的三个阶段中的每一个估计一个回归（这意味着我们在每个三个阶段的时间窗口结束时切割估计样本）。关于方程规格的更多细节见支持信息附录 C。支持信息附录 C 中的表 C1 和 C2 报告的回归结果显示，估计的相关性与关于净收入主要组成部分与风险加权资产（RWA）和一级资本比率之间关系的先前预期一致。特别是，我们发现净利息收入和其他收入的系数为正，而对家庭和企业的贷款的系数为负。所有变量和阶段的系数在统计上都是显著的，唯一的例外是巴塞尔 3 阶段对非金融公司的贷款。

Second, we check that the components of net income (net interest income; other income; operational expenses; loan loss provisions), default rates (of both NFCs and households), and stock prices have predictive power for a block of the other (macrofinancial and credit) variables included in the model, including the Tier 1 ratio. The aim of this exercise is to establish that our rich modeling of the banking sector helps improve the accuracy of predictability of Tier 1 relative to a “core” block of variables that are typically used in the literature (see, among others, Kanngiesser et al. 2020; Gambetti and Musso 2017). More in detail, we run a block-exogeneity test in which an “extended VAR model”, that is, including the nine core variables plus one additional variable at a time, is compared to the “restricted” VAR, that is, including only the nine core variables.

^{35}

For each variable, the test is run over four different samples: three up to the end of the time window of each of the three episodes (Basel III, 2011 EBA stress test, SSM/CA) plus one for the full sample (1993:Q1-2015:Q4). The results, reported in Table C3 in the Supporting Information Appendix C, suggest that, in almost all cases, each of the additional variables considered is helpful in forecasting the variables in the core VAR, including the Tier 1 ratio.
其次，我们检查净收入的组成部分（净利息收入；其他收入；运营费用；贷款损失准备金）、违约率（包括非金融公司和家庭）以及股票价格对模型中包含的其他（宏观金融和信贷）变量的预测能力，包括一级资本比率。此项工作的目的是确定我们对银行业的丰富建模有助于提高一级资本的可预测性，相较于文献中通常使用的“核心”变量块（参见，其他文献，Kanngiesser 等，2020；Gambetti 和 Musso，2017）。更详细地说，我们进行了一项块外生性检验，其中“扩展 VAR 模型”，即包括九个核心变量加上一个额外变量，与“限制性”VAR 进行比较，即仅包括九个核心变量。

^{35}

对于每个变量，测试在四个不同的样本上进行：三个样本截至每三个事件的时间窗口结束（巴塞尔协议 III，2011 年 EBA 压力测试，SSM/CA），加上一个完整样本（1993 年第一季度至 2015 年第四季度）。结果在支持信息附录 C 的表 C3 中报告，表明在几乎所有情况下，考虑的每个额外变量都有助于预测核心 VAR 中的变量，包括一级资本比率。

Third, we run a formal forecasting exercise over different time windows, in order to get a quantitative sense of the goodness-of-fit of the baseline model for Tier 1 ratio. In particular, we focus on time windows that differ from the three episodes that we consider to verify the effects of raising bank capital requirements: 1993:Q1-2006:Q4, a period ending well before the Global Financial Crisis and the beginning of the upward trend in bank capital and 1993:Q1-2012:Q1, which is the same period used for the “placebo” test (see Section 8). We use the root mean square forecast error to compare our baseline 16 -variable model to an

AR (1)

for Tier 1 ratio and to a BVAR including 9 variables (real GDP, consumer prices, shortand long-term interest rate, loan volumes, and rates to NFC and HH , and finally, the Tier 1 ratio). We consider horizons

H = 1, H = 4, H = 8

. The results, reported in Table C4 in the Supporting Information Appendix C, show that the baseline VAR model always outperforms the competing models, with the exception of the eight-quarter-ahead forecast in the first considered sample, where the baseline BVAR has a slightly higher RMSFE than the AR(1) model.
第三，我们在不同的时间窗口进行正式的预测练习，以量化基准模型对一级资本比率的拟合优度。特别是，我们关注与我们考虑的提高银行资本要求的三个阶段不同的时间窗口：1993 年第一季度至 2006 年第四季度，这是一个在全球金融危机之前结束的时期，以及 1993 年第一季度至 2012 年第一季度，这是用于“安慰剂”测试的相同时期（见第 8 节）。我们使用均方根预测误差将我们的基准 16 变量模型与一级资本比率的

AR (1)

进行比较，并与包括 9 个变量的 BVAR 进行比较（实际 GDP、消费价格、短期和长期利率、贷款量，以及对非金融公司和家庭的利率，最后是一级资本比率）。我们考虑的时间范围为

H = 1, H = 4, H = 8

。结果在附录 C 的支持信息表 C4 中报告，显示基准 VAR 模型始终优于竞争模型，唯一的例外是第一个考虑样本中的八个季度前预测，其中基准 BVAR 的均方根预测误差略高于 AR(1)模型。

Overall, the evidence stemming from all the exercises is reassuring about the adequacy of our estimated BVAR for modeling bank capital ratio. Armed with this tool, we turn to defining our empirical approach for the scenario analysis.
总体而言，所有练习所提供的证据令人放心，我们估计的 BVAR 在建模银行资本比率方面是足够的。借助这个工具，我们开始定义情景分析的实证方法。

4.5 | Conditional Forecasting and Scenario Analysis
4.5 | 条件预测与情景分析

As explained in the next Section (Section 5), we use conditional forecasting methodology to produce one counterfactual of the Tier 1 ratio (counterfactual B). This procedure is called “conditional-on-variables”

^{36}

; it relies on the reduced-form parameters of the VAR only and does not require to recover the the structural form (Waggoner and Zha 1999; Antolín-Díaz, Petrella, and Rubio-Ramírez 2021).
如下一节（第 5 节）所述，我们使用条件预测方法生成一个一级资本比率的反事实（反事实 B）。该程序称为“条件变量”

^{36}

；它仅依赖于 VAR 的简化形式参数，不需要恢复结构形式（Waggoner 和 Zha 1999；Antolín-Díaz、Petrella 和 Rubio-Ramírez 2021）。

We adapt this methodology by using expectations instead of actual variables in the conditioning set. Conditional forecasts are produced using the standard algorithm in the VAR literature developed by Doan, Litterman, and Sims (1986), which consists of solving a least squares problem to pick the shocks needed to satisfy the conditions. The conditional forecasts are obtained by imposing a dynamic pattern of residuals

ε_{t}

compatible with the required conditions on observables for a given desired horizon H. As proved by Waggoner and Zha (1999), the distribution of such conditional forecasts is invariant to an orthonormal transformation of the underlying factorization of the covariance matrix of the residuals, which is therefore assumed to be triangular without any loss of generality. In the implementation we form the posterior distribution of VAR parameters without taking into account the conditions to be imposed.

^{37}

For each model, we use Monte Carlo simulations to obtain draws of the BVAR coefficients and the error variance matrix from the standard posterior. In the case of models with time-varying coefficients and time-varying volatility we hold the various parameters and volatilities constant at their end-of-sample estimation values over the 2-year forecast horizon to simplify the computation of conditional forecasts.
我们通过在条件集中的使用预期而不是实际变量来调整这种方法论。条件预测是使用 Doan、Litterman 和 Sims（1986）开发的 VAR 文献中的标准算法生成的，该算法包括解决一个最小二乘问题，以选择满足条件所需的冲击。条件预测是通过施加与给定期望时间 H 的可观测条件兼容的残差动态模式

ε_{t}

来获得的。正如 Waggoner 和 Zha（1999）所证明的，这种条件预测的分布对残差协方差矩阵的基础分解的正交变换是不变的，因此假设其为三角形而不失一般性。在实施中，我们在不考虑要施加的条件的情况下形成 VAR 参数的后验分布。

^{37}

对于每个模型，我们使用蒙特卡洛模拟从标准后验中获得 BVAR 系数和误差方差矩阵的抽样。在具有时间变化系数和时间变化波动率的模型中，我们在 2 年预测期内将各种参数和波动率保持在其样本结束估计值不变，以简化条件预测的计算。

5 | Recovering the Impact of Regulatory and Supervisory Actions
5 | 恢复监管和监督行动的影响

In this section, we describe the methodology used to recover the size of bank capital increases associated with supervisory/ regulatory initiatives and to study the associated evolution of lending supply and economic activity. The procedure relies on scenario analysis and borrows, especially, from the literature on non-standard monetary policy (Altavilla, Canova, and Ciccarelli 2020; Dahlhaus, Hess, and Reza 2018; Lenza, Pill, and Reichlin 2010; Giannone et al. 2012; Altavilla, Giannone, and Lenza 2016).
在本节中，我们描述了用于恢复与监管/监管举措相关的银行资本增加规模的方法论，并研究了与之相关的贷款供应和经济活动的演变。该程序依赖于情景分析，特别借鉴了关于非常规货币政策的文献（Altavilla, Canova, and Ciccarelli 2020; Dahlhaus, Hess, and Reza 2018; Lenza, Pill, and Reichlin 2010; Giannone et al. 2012; Altavilla, Giannone, and Lenza 2016）。

5.1 | Step 1: Estimating the "Size" of the Policy Actions
5.1 | 第一步：估算政策行动的“规模”

For each of the three time-windows described in Section 3, we first obtain counterfactual series of the Tier 1 ratio, which represents the path that capital would have followed if no policy actions had been taken. As described below, this counterfactual will be then compared to the actual development of the Tier 1 ratio, retrieving a measure of the size of the policy actions.
对于第 3 节中描述的三个时间窗口，我们首先获得一级资本充足率的反事实系列，这代表了如果没有采取政策行动，资本将遵循的路径。如下面所述，这个反事实将与一级资本充足率的实际发展进行比较，从而获取政策行动规模的衡量。

More in detail, we compute counterfactuals based on two different strategies:
更详细地说，我们基于两种不同的策略计算反事实
A. Counterfactual as unconditional BVAR forecast. We let the Tier 1 ratio evolve unconditionally according to its BVAR forecast over the aftermath of each episode of interest (Basel III, the EBA stress test, and the launch of the SSM and the ECB CA). Denoting the counterfactual Tier 1 ratio by

{\hat{k}}_{t}

, we formally assume that

{\hat{k}}_{1, t + h} = E (Y_{t})

h = 1, 2, \dots, 8

. This is the strategy followed, for example, by Dahlhaus, Hess, and Reza (2018) to estimate the effects of the Federal Reserve balance sheet expansion after the Lehman failure and by Altavilla, Canova, and Ciccarelli (2020) to assess the impact of ECB’s negative deposit facility rate and the Asset Purchases Program on the EONIA rate, sovereign bond yields and bank bond yields. The unconditional path of the Tier 1 ratio can be interpreted as the path that may have been expected based on past regularities in the economy.
A. 反事实作为无条件 BVAR 预测。我们让一级资本充足率根据其 BVAR 预测在每个感兴趣事件（巴塞尔协议 III、EBA 压力测试、SSM 的启动和 ECB CA）之后无条件演变。用

{\hat{k}}_{t}

表示反事实一级资本充足率，我们正式假设

{\hat{k}}_{1, t + h} = E (Y_{t})

，

h = 1, 2, \dots, 8

。这就是 Dahlhaus、Hess 和 Reza（2018）在估计雷曼破产后美联储资产负债表扩张的影响时所遵循的策略，以及 Altavilla、Canova 和 Ciccarelli（2020）在评估 ECB 负存款便利利率和资产购买计划对 EONIA 利率、主权债券收益率和银行债券收益率的影响时所遵循的策略。一级资本充足率的无条件路径可以解释为基于经济过去规律可能预期的路径。
B. Counterfactual as conditional BVAR forecast. We use the estimates obtained until the last available point in the subsample to compute a forecast of the Tier 1 ratio conditional on the real-time expected path of the main determinants of bank capital. Based on the considerations discussed in Section 4.4, in the conditioning set (for the baseline exercises), we include the following: real GDP, inflation, the short-term policy rate, the Italian 10-year sovereign yield (which captures both changes in the long-term free interest rate and controls for tension on the sovereign debt market), one variable measuring bank profitability (net interest income in the baseline), and firms’ and households’ default rates.

^{38}

The path of the variables in the conditioning set is the one consistent with the latest available vintage of the Eurosystem MPE before the end of the estimation sample (for each of the three episodes considered; for example, for the SSM/CA episode, we take the Winter Forecast of December 2013). These projections-which are formulated at quarterly frequency and are restricted to authorized Eurosystem staff—represent (the best available) real-time forecast of the key drivers of bank capital. In practice, we collect the forecast made by the Eurosystem Staff at each data point consistent with our exercise

{\hat{k}}_{2, t + h} = E (Y_{t}, x_{t + h}), h = 1, 2, \dots, 8

. This counterfactual Tier 1 ratio can be considered as the expected path of Tier 1 based on the exogenous drivers that were predictable in real time. This approach is similar to the one proposed by Lenza, Pill, and Reichlin (2010) and Giannone et al. (2012), among others, to recover the size of the Eurosystem balance sheet in the aftermath of the Global Financial Crisis. One important difference is, however, that they rely on the actual values of the

x_{t + h}

; while this choice is justified when considering a policy variable under the strict control of the Central Bank (its balance-sheet), it would pose a potential contradiction if applied to the Tier 1 ratio.

^{39}

In addition, using expecta-tions-instead of actual values-has the advantage of delivering a quasi-real-time assessment of developments in Tier 1 ratio, which is useful from a policy perspective.
B. 反事实作为条件 BVAR 预测。我们使用在子样本中最后可用点获得的估计值来计算基于银行资本主要决定因素实时预期路径的一级资本比率预测。基于第 4.4 节讨论的考虑，在条件集（基线练习）中，我们包括以下内容：实际 GDP、通货膨胀、短期政策利率、意大利 10 年期主权收益率（该收益率既反映长期无风险利率的变化，又控制主权债务市场的紧张情况）、一个衡量银行盈利能力的变量（基线中的净利息收入）以及企业和家庭的违约率。

^{38}

条件集中变量的路径与估计样本结束前最新可用版本的欧洲系统 MPE 一致（对于考虑的三个事件；例如，对于 SSM/CA 事件，我们采用 2013 年 12 月的冬季预测）。这些预测——以季度频率制定，仅限于授权的欧洲系统工作人员——代表了（最佳可用的）银行资本关键驱动因素的实时预测。实际上，我们收集欧洲系统工作人员在与我们练习一致的每个数据点所做的预测

{\hat{k}}_{2, t + h} = E (Y_{t}, x_{t + h}), h = 1, 2, \dots, 8

。这个反事实的一级资本比率可以被视为基于实时可预测的外生驱动因素的一级资本预期路径。这种方法类似于 Lenza、Pill 和 Reichlin（2010）以及 Giannone 等人（2012）提出的方法，用于恢复全球金融危机后欧洲系统资产负债表的规模。然而，一个重要的区别是，他们依赖于

x_{t + h}

的实际值；虽然在考虑中央银行严格控制的政策变量（其资产负债表）时，这种选择是合理的，但如果应用于一级资本比率，则可能会产生潜在的矛盾。

^{39}

此外，使用预期——而不是实际值——的优点在于能够提供对一级资本比率发展的准实时评估，这在政策角度上是有用的。

Once we have obtained these counterfactual paths, we compare them with the observed evolution in the ratio during each time window. By construction, the two counterfactuals series of the Tier 1 are not affected by the policy actions undertaken during the three time windows under consideration. On the contrary, the realized path of capital includes these effects. Thus, following Dahlhaus, Hess, and Reza (2018), Altavilla, Canova, and Ciccarelli (2020), Lenza, Pill, and Reichlin (2010) and Giannone et al. (2012), we can loosely interpret the difference between actual and the counterfactual capital ratios in as a measure of the size of the policy actions that occurred during each time window. A discussion of this interpretation is provided in Section 5.3 below.
一旦我们获得了这些反事实路径，我们将其与每个时间窗口内观察到的比率演变进行比较。根据构造，Tier 1 的两个反事实系列不受考虑的三个时间窗口内采取的政策行动的影响。相反，资本的实际路径包括这些影响。因此，遵循 Dahlhaus、Hess 和 Reza（2018）、Altavilla、Canova 和 Ciccarelli（2020）、Lenza、Pill 和 Reichlin（2010）以及 Giannone 等人（2012）的研究，我们可以大致将实际资本比率与反事实资本比率之间的差异解释为在每个时间窗口内发生的政策行动的规模的度量。关于这种解释的讨论在下面的第 5.3 节中提供。

5.2 | Step 2: Estimating the Impact of the Policy Actions on Lending Supply and Economic Activity
5.2 | 第 2 步：估计政策措施对贷款供应和经济活动的影响

Once we have retrieved the size of the regulatory/supervisory actions in each of the three windows, we can compute the associated evolution of real GDP, consumer prices, and the cost and availability of bank credit. In order to do so, for each of these variables, we compute the difference between two conditional forecasts:
一旦我们获取了三个窗口中监管/监督行动的规模，我们就可以计算实际 GDP、消费者价格以及银行信贷的成本和可获得性的相关演变。为此，对于这些变量中的每一个，我们计算两个条件预测之间的差异：
(i) a forecast conditional on the actual path of the Tier 1 ratio (policy scenario), and
(i) 基于一级资本比率实际路径的预测（政策情景），并
(ii) a forecast conditional on each of the two (counterfactual) paths for bank capital described above in Step 1 (no-policy scenario).
(ii) 基于步骤 1 中描述的两条（反事实）银行资本路径的预测（无政策情景）。

We refer to the two scenarios as the policy and no-policy scenarios, respectively. The reason for this is that the first scenario is based on the evolution of the Tier 1 ratio which by construction includes the policy initiatives, while the second is based on the path that the Tier 1 would have had, had those policy initiatives not been taken. These two simulations mainly differ with respect to the evolution of the capital ratio: consequently, this procedure is (loosely) equivalent to computing impulse responses to the shock of interest (see Dahlhaus, Hess, and Reza 2018; Altavilla, Canova, and Ciccarelli 2020; Lenza, Pill, and Reichlin 2010; Giannone et al. 2012; Jardet, Monfort, and Pegoraro 2013; Banbura, Giannone, and Lenza 2015). More in detail, as shown by Lenza, Pill, and Reichlin (2010), for each endogenous variable

Y_{i}

, the policy scenario is the conditional expectation based on the estimated parameters

A (L)

, the past and current values

Y_{i 0}, Y_{i 1}, \dots, Y_{i t}

and the actual values of the Tier 1 capital:
我们将这两种情景分别称为政策情景和无政策情景。原因在于，第一个情景是基于一级资本比率的演变，而该比率的构成包括政策举措，而第二个情景则是基于如果没有采取这些政策举措，一级资本比率将会有的路径。这两种模拟主要在资本比率的演变上有所不同：因此，这一过程（大致上）等同于计算对感兴趣冲击的脉冲响应（参见 Dahlhaus, Hess, 和 Reza 2018；Altavilla, Canova, 和 Ciccarelli 2020；Lenza, Pill, 和 Reichlin 2010；Giannone 等 2012；Jardet, Monfort, 和 Pegoraro 2013；Banbura, Giannone, 和 Lenza 2015）。更详细地说，如 Lenza, Pill, 和 Reichlin（2010）所示，对于每个内生变量

Y_{i}

，政策情景是基于估计参数

A (L)

、过去和当前值

Y_{i 0}, Y_{i 1}, \dots, Y_{i t}

以及一级资本的实际值的条件期望：

E_{(A (L))} (Y_{i t} ∣ Y_{i 0}, Y_{i 1}, \dots, Y_{i t}, k_{P, t + 1}, \dots, k_{P, t + H})

where

H

is the forecast horizon (

H = 8

in our analysis). The nopolicy scenario is obtained as above replacing the actual value of the Tier 1 with its counterfactual path

{\hat{k}}_{s, N P, t + H}

, where

s = 1, 2

denotes the strategy-the unconditional or real-time conditional forecast-obtained in Step 1:
其中

H

是预测期限（在我们的分析中为

H = 8

）。无政策情景是通过上述方法获得的，将一级资本的实际值替换为其反事实路径

{\hat{k}}_{s, N P, t + H}

，其中

s = 1, 2

表示在步骤 1 中获得的策略——无条件或实时条件预测：

E_{(A (L))} (Y_{i t} ∣ Y_{i 0}, Y_{i 1}, \dots, Y_{i t}, {\hat{k}}_{s, N P, t + 1}, \dots, {\hat{k}}_{s, N P, t + H})

The impact of the policy actions is thus computed, for each variable

Y_{i}

, as the difference between (7) and (8).
因此，政策行动的影响被计算为每个变量

Y_{i}

的 (7) 和 (8) 之间的差异。

Importantly, when running these simulations, we impose two additional conditions:
重要的是，在运行这些模拟时，我们施加了两个额外的条件：
(i) that the path of monetary policy is the same between the two scenarios over the entire simulation horizon;
(i) 在整个模拟期内，货币政策的路径在两种情景下是相同的；
(ii) that the values of the slow-moving variables, that is, real GDP, HICP, and loans to NFCs and HHs, are the same across the two scenarios on the first period of the simulations.

^{40}

(ii) 在模拟的第一期中，缓慢变化变量的值，即实际 GDP、HICP 以及对非金融公司和家庭的贷款，在两个情景中是相同的。

^{40}

The latter condition is equivalent to imposing that these variables do not react on impact. In the next section, we will discuss why imposing these conditions is important for the interpretation of our results.
后一个条件等同于要求这些变量在冲击下不发生反应。在下一节中，我们将讨论为什么施加这些条件对我们结果的解释是重要的。

5.3 | Discussion5.3 | 讨论

A crucial issue in the empirical strategy described above is whether the difference between the two scenarios correctly measures the size of the policy actions and their impact on the macroeconomic and credit variables. In the previous sections, we have provided a number of compelling arguments why it is reasonable to think that this is the case. Further analyses and robustness checks presented in the following sections will strengthen this claim. All these arguments can be summarized as follows:
上述经验策略中的一个关键问题是，两个情景之间的差异是否正确衡量了政策行动的规模及其对宏观经济和信贷变量的影响。在前面的部分中，我们提供了一些令人信服的论据，说明认为情况确实如此是合理的。接下来的部分中将呈现进一步的分析和稳健性检验，以加强这一论点。所有这些论点可以总结如下：

Our interpretation is supported by the compelling narrative during the event windows. As we explained in Section 3, in all three episodes, either the authorities explicitly asked Italian banks to significantly increase their capital levels, as a consequence of increased requirements or shortfalls identified in the stress tests or the CA, or banks increased their capitalization in advance of the stress tests or the CA, in anticipation of-and with the objective of avoiding-potential shortfalls.
我们的解释得到了事件窗口期间引人注目的叙述的支持。正如我们在第 3 节中所解释的，在所有三个事件中，或者当局明确要求意大利银行显著提高其资本水平，这是由于压力测试或 CA 中识别出的要求增加或短缺，或者银行在压力测试或 CA 之前提前增加了其资本化，以预期并旨在避免潜在的短缺。
The rich characterization of our model takes into account the potential interactions between bank capital and all its main drivers. In Section 4.4, we have discussed theoretical and empirical reasons why the variables included in our model have high predictive power for the Tier 1 ratio.
我们模型的丰富特征考虑了银行资本与其所有主要驱动因素之间的潜在互动。在第 4.4 节中，我们讨论了为什么我们模型中包含的变量对一级资本比率具有很高的预测能力的理论和实证原因。
The restrictions imposed on the behavior of monetary policy and the other slow-moving variables in the simulations within our second step (see Section 5.2) significantly restrict the combinations of shocks that may be used to satisfy the conditioning path; they are equivalent to assuming that policy/regulatory actions were the main driver of changes of the Tier 1 ratio on impact (Rostagno et al. 2019).
在我们第二步的模拟中，对货币政策行为和其他缓慢变化变量施加的限制（见第 5.2 节）显著限制了可用于满足条件路径的冲击组合；这相当于假设政策/监管行动是影响一级资本比率变化的主要驱动因素（Rostagno et al. 2019）。
Moreover, the results obtained with the scenario analysis are very similar to those obtained using a structural VAR model in which exogenous shocks to bank capital requirements are identified using the narrative sign restrictions methodology developed by Antolín-Díaz and Rubio-Ramírez (2018) (see our companion paper Conti, Nobili, and Signoretti 2023). This confirms that the effects estimated in the scenario analysis capture the impact of an increase in Tier 1 ratio due to exogenous factors, even though we are modeling here a persistent shock-rather than a transitory one.
此外，情景分析获得的结果与使用结构性 VAR 模型获得的结果非常相似，该模型中通过 Antolín-Díaz 和 Rubio-Ramírez（2018）开发的叙述性符号限制方法识别对银行资本要求的外生冲击（参见我们伴随的论文 Conti, Nobili, 和 Signoretti 2023）。这确认了情景分析中估计的影响捕捉了由于外生因素导致的一级资本比率增加的影响，即使我们在这里建模的是一个持久性冲击，而不是一个暂时性冲击。
In Section 8, we perform several additional analyses and robustness exercises, all of which confirm that the results obtained in the scenario analysis are correctly attributed to be the effect of increased regulatory/supervisory pressure on bank capital. Among these, we will discuss complementary evidence available from the ECB BLS which strongly supports the notion that the increases in banks’ capital ratios during the EBA and the SSM/CA episodes were driven by regulatory/supervisory actions; we will show that the actual developments of the main variables of interest fall within the credibility interval of their (conditional) forecasts in the policy scenario, confirming that no other major events occurred in the windows and reinforcing our case that the dynamics of the Tier 1 ratio may be attributed to the actions of the regulators/supervisors; we will confirm our results in a model that includes the US Tier 1 ratio as an additional endogenous variable, capturing a global banking factor and thus helping to isolate the evolution of the Italian Tier 1 ratio from drivers unrelated to the “European” policy actions; in a “placebo” test, we will verify that the scenario analysis correctly predicts the Tier 1 ratio in periods outside our policy event windows, thus showing the absence of systematic deviations of the actual Tier 1 ratio from its expected path; finally, we will separately include the numerator (Tier 1 equity) and the denominator (RWAs) of the Tier 1 ratio in the VAR, confirming our results and providing hints on potential differences in transmission channels to the real economy across the regulatory/supervisory episodes.
在第 8 节中，我们进行了几项额外的分析和稳健性检验，所有这些都确认在情景分析中获得的结果确实可以归因于监管/监督压力增加对银行资本的影响。在这些内容中，我们将讨论来自欧洲央行 BLS 的补充证据，这些证据强烈支持银行资本比率在 EBA 和 SSM/CA 阶段的增加是由监管/监督行动驱动的；我们将展示主要关注变量的实际发展落在其（条件）预测的可信区间内，确认在这些窗口期没有发生其他重大事件，并强化我们的观点，即一级资本比率的动态可能归因于监管者/监督者的行动；我们将在一个模型中确认我们的结果，该模型将美国一级资本比率作为一个额外的内生变量，捕捉全球银行因素，从而帮助将意大利一级资本比率的演变与与“欧洲”政策行动无关的驱动因素隔离；在一个“安慰剂”测试中，我们将验证情景分析在我们的政策事件窗口之外的时期正确预测一级资本比率，从而显示实际一级资本比率与其预期路径之间没有系统性偏差；最后，我们将在 VAR 中单独包含一级资本比率的分子（一级资本）和分母（风险加权资产），确认我们的结果并提供关于监管/监督阶段对实际经济传导渠道潜在差异的线索。

6 | Baseline Results
6 | 基准结果

The first row of Figures 2-4 reports, for each episode, (i) the actual values of the banks’ Tier 1 capital ratio (the black lines) against its counterfactual dynamics, obtained according to both of the strategies described in Section 5 and (ii) the correspondent estimated size of the policy actions, computed as the difference between the actual and the counterfactual. Accordingly, all charts report two solid lines: the blue line, for the case in which the Tier 1 ratio counterfactual is its BVAR unconditional forecast (case A in Section 5.2), and the magenta line, for the case in which the counterfactual is obtained by the conditional forecast (case B). These lines are the median of the empirical posterior distribution of the out-of-sample conditional forecasts obtained by Monte Carlo simulations based on 1000 draws. The chart also reports the 0.68 and 0.90 probability intervals (the dark and light gray shaded areas, which refer to the blue solid line; and the dashed and dotted magenta lines, for the magenta line). The remaining rows of the figures report the percentage or absolute (in basis points; depending on the variable under consideration) difference between the level of real GDP, consumer prices, loans, and loan margins (defined as the difference between loan rates and the short-term policy rate) in the policy scenario relative to the no-policy scenario. As discussed in Section 5, these graphs can be loosely interpreted as IRFs. Results for the three windows are also summarized in Table 1.
图 2-4 的第一行报告了每个事件的(i) 银行的一级资本充足率的实际值（黑线）与其反事实动态的对比，这些动态是根据第 5 节中描述的两种策略获得的，以及(ii) 政策行动的相应估计规模，计算为实际值与反事实之间的差异。因此，所有图表都报告了两条实线：蓝线表示一级资本充足率反事实为其 BVAR 无条件预测的情况（第 5.2 节中的案例 A），而品红线则表示反事实是通过条件预测获得的情况（案例 B）。这些线是基于 1000 次抽样的蒙特卡洛模拟获得的样本外条件预测的经验后验分布的中位数。图表还报告了 0.68 和 0.90 的概率区间（深灰色和浅灰色阴影区域，分别对应蓝色实线；以及品红线的虚线和点线）。其余行的数字报告了政策情景与无政策情景下实际 GDP、消费价格、贷款和贷款利差（定义为贷款利率与短期政策利率之间的差异）水平之间的百分比或绝对（以基点计；取决于所考虑的变量）差异。如第 5 节所讨论，这些图表可以大致解释为 IRFs。三种窗口的结果也在表 1 中进行了总结。

Our results show that, in the Basel 3 episode, the increase in the Tier 1 ratio at the end of the 2 -year window is correctly predicted by our BVAR, meaning that bank capital increases were consistent with macro and banking developments and signaling the good fitting properties of the model. Consistently, no effect on lending or macroeconomic variables is found during this episode.
我们的结果表明，在巴塞尔 3 时期，2 年窗口期末一级资本比率的增加得到了我们的 BVAR 的正确预测，这意味着银行资本的增加与宏观和银行发展是一致的，并且表明模型的良好拟合特性。在此期间，未发现对贷款或宏观经济变量的影响。

Instead, both in the 2011 EBA stress test and in the 2014 SSM/ CA episodes, the actual dynamics of the Tier 1 ratio lies outside the credibility interval of both the unconditional and the conditional forecasts. At the end of the 2-year horizon, the estimated size of the policy actions in the two episodes is 2.6 and 2.3 percentage points. The behavior of the main macroeconomic variables suggests that these capital increases led to a tightening of credit supply conditions: On average over the two periods, the stocks of loans to NFCs were lower by between

4.4 %

and

5.2 %

(at the end of the horizon, depending on the episode considered and whether we consider the counterfactual based on the unconditional forecast or the conditional). The stock of loans to households declined significantly in the SSM/CA episodes, by between

4.4 %

and

5.5 %

; in the EBA episode; instead, the credibility interval of the IRF both simulations includes the zero, suggesting that no significant impact is detected. Loan interest rate margins (i.e., the difference between loan rates and the short-term interest rate) increased by between 80 and 130 basis points and by between 100 and 150 bps, respectively, for NFCs and HHs. As a consequence of the restriction in credit supply conditions, GDP was significantly affected: During the EBA episode, the decline was concentrated in the first year following the policy action (with a peak effect of about

2 %

); in the SSM/CA episode, the largest impact is estimated at the end of the horizon, with an impact between

3 %

and

4 %

(depending on the counterfactual considered). Finally, the reaction of consumer prices was heterogeneous, with a decline in the EBA episode and an increase in the SSM/CA episode.

相反，在 2011 年 EBA 压力测试和 2014 年 SSM/CA 事件中，一级资本比率的实际动态超出了无条件和有条件预测的可信区间。在 2 年期末，两个事件中政策行动的估计规模分别为 2.6 和 2.3 个百分点。主要宏观经济变量的行为表明，这些资本增加导致了信贷供应条件的收紧：在两个时期的平均水平上，对非金融公司的贷款存量降低了

4.4 %

到

5.2 %

（在考虑的事件和我们是否考虑基于无条件预测或有条件预测的反事实的情况下，在期末）。在 SSM/CA 事件中，家庭贷款的存量显著下降，降幅在

4.4 %

到

5.5 %

之间；而在 EBA 事件中，两个模拟的脉冲响应函数的可信区间均包括零，表明未检测到显著影响。贷款利率差（即贷款利率与短期利率之间的差额）分别增加了 80 到 130 个基点和 100 到 150 个基点，针对非金融公司（NFCs）和家庭（HHs）。由于信贷供应条件的限制，GDP 受到了显著影响：在 EBA 事件期间，下降主要集中在政策行动后的第一年（峰值效应约为

2 %

）；在 SSM/CA 事件中，预计最大影响发生在时间范围的末尾，影响在

3 %

和

4 %

之间（取决于考虑的反事实）。最后，消费者价格的反应是异质的，在 EBA 事件中下降，而在 SSM/CA 事件中上升。

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of unconditional forecasts, obtained by a simulation of 5,000 draws; the straight magenta line is the policy variable out-of-sample VAR conditional forecast on MPE values of real GDP, HICP the short- and long-term rate, net-interest income and default rates of NFC and HH while the dashed (dotted) magenta lines represent

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of conditional forecasts, obtained by a simulation of 5,000 draws. Estimation sample is 1993:Q1 - 2009:Q1. Second-Eighth panel: the blue line is the difference between the policy and no policy scenario defined in terms of the unconditional forecast and the dark (light) grey shaded area represents

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of this difference; the magenta line is the difference (either percentage or basis points) between the policy and no policy scenario defined in terms of the conditional forecast and the dashed (dotted) magenta lines represent

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of conditional forecasts. Forecasting sample is 2009:Q2 - 2010:Q4.
备注：第一面板：黑线是政策变量（一级资本比率）的实际数据；蓝线是其样本外 VAR 无条件预测，深（浅）灰色阴影区域是

16 - 84 % (5 - 95 %)

百分位数的无条件预测经验后验分布，通过 5000 次抽样模拟获得；直的品红色线是基于实际 GDP、HICP 短期和长期利率、净利息收入以及非金融公司和家庭的违约率的政策变量样本外 VAR 条件预测，而虚线（点线）品红色线代表

16 - 84 % (5 - 95 %)

百分位数的条件预测经验后验分布，通过 5000 次抽样模拟获得。估计样本为 1993 年第一季度至 2009 年第一季度。第二至第八面板：蓝线是政策与无政策情景之间的差异，定义为无条件预测，深灰色（浅灰色）阴影区域表示

16 - 84 % (5 - 95 %)

的经验后验分布的百分位数；品红色线是政策与无政策情景之间的差异（以百分比或基点表示），定义为有条件预测，虚线（点线）品红色线表示

16 - 84 %

（

5 - 95 %

）的有条件预测的经验后验分布的百分位数。预测样本为 2009 年第二季度至 2010 年第四季度。

FIGURE 2 | The effects of increasing bank capital requirements: Basel III episode.
图 2 | 提高银行资本要求的影响：巴塞尔 III 事件。

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of conditional forecasts, obtained by a simulation of 5,000 draws. Estimation sample is 1993:Q1 - 2010:Q4. Second-Eighth panel: the blue line is the difference between the policy and no policy scenario defined in terms of the unconditional forecast and the dark (light) grey shaded area represents

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of this difference; the magenta line is the difference (either percentage or basis points) between the policy and no policy scenario defined in terms of the conditional forecast and the dashed (dotted) magenta lines represent

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of conditional forecasts. Forecasting sample is 2011:Q1 - 2012:Q4
备注：第一面板：黑线是政策变量（一级资本比率）的实际数据；蓝线是其样本外 VAR 无条件预测，深（浅）灰色阴影区域是

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

百分位数的条件预测经验后验分布，通过 5000 次抽样模拟获得。估计样本为 1993 年第一季度至 2010 年第四季度。第二至第八面板：蓝线表示政策与无政策情景之间的差异，定义为无条件预测，深灰色（浅灰色）阴影区域代表

16 - 84 %

（

5 - 95 %

）的经验后验分布的百分位数；品红色线表示政策与无政策情景之间的差异（以百分比或基点表示），定义为有条件预测，虚线（点线）品红色线代表

16 - 84 % (5 - 95 %)

的有条件预测的经验后验分布的百分位数。预测样本为 2011 年第一季度至 2012 年第四季度。

FIGURE 3 | The effects of increasing bank capital requirements: EBA episode.
图 3 | 提高银行资本要求的影响：EBA 事件。

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of conditional forecasts, obtained by a simulation of 5,000 draws. Estimation sample is 1993:Q1 - 2013:Q4. Second-Eighth panel: the blue line is the difference between the policy and no policy scenario defined in terms of the unconditional forecast and the dark (light) grey shaded area represents

16 - 84 %

(

5 - 95 %

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of conditional forecasts. Forecasting sample is 2014:Q1 - 2015:Q4
备注：第一面板：黑线是政策变量（一级资本比率）的实际数据；蓝线是其样本外 VAR 无条件预测，深（浅）灰色阴影区域是

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

百分位数的条件预测经验后验分布，通过 5000 次抽样模拟获得。估计样本为 1993 年第一季度至 2013 年第四季度。第二至第八面板：蓝线是政策与无政策情景之间的差异，定义为无条件预测，深灰色（浅灰色）阴影区域代表

16 - 84 %

（

5 - 95 %

）的经验后验分布的百分位数；品红色线是政策与无政策情景之间的差异（以百分比或基点表示），定义为有条件预测，虚线（点线）品红色线代表

16 - 84 %

（

5 - 95 %

）的有条件预测的经验后验分布的百分位数。预测样本为 2014 年第一季度至 2015 年第四季度。

FIGURE 4 | The effects of increasing bank capital requirements: SSM episode.
图 4 | 提高银行资本要求的影响：SSM 事件。
TABLE 1 | Cross-episodes effects of bank capital requirements increases in the fixed-coefficients BVAR.
表 1 | 固定系数 BVAR 中银行资本要求增加的跨期效应。

Episode (2-year window)集 (2 年窗口)

Horizon地平线

政策行动的预计规模（百分点）一级资本充足率

Estimated size of the policy actions (pp)

Tier 1 ratio

Estimated impact after

H

periods
预计在

H

个周期后的影响

Macro variables宏观变量

Loan volume贷款金额

Loan margin (bps)贷款利差（基点）

Real GDP实际国内生产总值

HICP

NFCs

HHs

NFCs

HHs

Unconditional无条件

EBA

0.371

0.000

12.477

-7.467

1.290

-0.018

0.001

-0.018

-0.004

56.235

53.014

2.605

-0.002

-0.009

-0.050

0.017

80.190

102.426

SSM/CA

0.224

0.000

4.805

-17.398

1.363

-0.019

0.005

-0.011

-0.014

43.585

36.412

2.336

-0.031

0.013

-0.052

-0.055

118.664

125.266

Average平均

0.298

0.000

8.641

-12.433

1.326

-0.019

0.003

-0.015

-0.009

49.910

44.713

2.471

-0.017

0.002

-0.051

-0.019

99.427

113.846

Conditional条件

EBA

0.384

0.000

2.332

7.899

1.444

-0.010

0.003

-0.015

-0.006

91.096

80.099

2.554

-0.004

0.002

-0.053

0.018

131.499

165.581

SSM/CA

0.191

0.000

10.694

-27.071

1.174

-0.019

0.004

-0.011

-0.021

18.620

27.718

2.207

-0.038

0.014

-0.044

129.580

152.417

Average平均

0.288

0.000

6.513

-9.586

1.309

-0.015

0.003

-0.013

54.858

53.909

2.380

-0.021

0.008

-0.049

-0.013

130.540

158.999

In terms of the economic mechanisms behind the impact of the policy actions on GDP, we have in mind the (long-dating) literature on the credit channel of monetary policy and, in particular, on the impact of credit supply restrictions on GDP (Bernanke and Gertler 1995). These effects may run both through increases in lending spreads, which affect aggregate demand and thus GDP and via direct quantity restrictions (rationing).
在政策行动对 GDP 影响的经济机制方面，我们考虑到关于货币政策信用渠道的（长期）文献，特别是关于信贷供应限制对 GDP 影响的研究（Bernanke and Gertler 1995）。这些影响可能通过贷款利差的增加影响总需求，从而影响 GDP，也可能通过直接的数量限制（配给）来实现。

Our estimated effects on GDP are in line with those results in the literature on the impact of bank capital shocks. In particular, we can compare our result to peak effects in the works by Meeks (2017), Mésonnier and Stevanovic (2017), Mésonnier and Monks (2015), and Kanngiesser et al. (2020), normalizing the magnitude of the increase in Tier 1 ratio to 2.5 pp (which is roughly the increase in capitalization that we derive in our EBA and SSM/CA episodes). The impact of GDP in these papers ranges between

1 %

and

3 %

; the decline in loans to NFCs is larger than the average effect in our simulations in three out of the four papers mentioned; the increase in loan spreads (considering both households and firms) ranges between 30 and 150 bps . In addition, one has to bear in mind that the cited literature is based on structural analysis and as such their results focus on averages across episodes; in contrast, our results are specific to episodes during which the narrative suggests that the size and the impact of the policy actions is likely to have been largest, which would, in principle, justify us finding potential larger effects.
我们对 GDP 的估计影响与文献中关于银行资本冲击影响的结果一致。特别是，我们可以将我们的结果与 Meeks (2017)、Mésonnier 和 Stevanovic (2017)、Mésonnier 和 Monks (2015)以及 Kanngiesser 等人 (2020) 的峰值效应进行比较，将一级资本比率的增加幅度标准化为 2.5 个百分点（这大致是我们在 EBA 和 SSM/CA 事件中得出的资本化增加幅度）。这些论文中 GDP 的影响范围在

1 %

到

3 %

之间；对非金融公司的贷款下降在提到的四篇论文中有三篇的平均效果更大；贷款利差的增加（考虑家庭和企业）范围在 30 到 150 个基点。此外，必须记住，所引用的文献基于结构分析，因此它们的结果侧重于各事件的平均值；相比之下，我们的结果特定于叙述表明政策行动的规模和影响可能最大的一些事件，这原则上可以证明我们发现潜在更大效应的合理性。

The impacts on loans and spreads are also of the same magnitude of those obtained-in a different setting-by DelGiovane, Nobili, and Signoretti (2017), who estimate the impact of the tightening of bank capital constraints during the Global and the Sovereign debt crisis, based on Italian banks’ replies to the BLS. Similarly, our results are also comparable to the responses of the US economy to a bank credit supply shock reported in Bassett, Lee, and Spiller (2015). There, a shock that produces a

4 %

decline in lending capacity (loans outstanding and unused commitments) raises corporate bond spreads by 40 basis points and causes a fall of up to

0.7 %

in real GDP.
对贷款和利差的影响与 DelGiovane、Nobili 和 Signoretti（2017）在不同背景下获得的结果相同，他们根据意大利银行对 BLS 的回复估计了全球和主权债务危机期间银行资本约束收紧的影响。同样，我们的结果也可以与 Bassett、Lee 和 Spiller（2015）报告的美国经济对银行信贷供应冲击的反应进行比较。在那里，产生

4 %

的贷款能力下降（未偿贷款和未使用的承诺）使企业债券利差上升 40 个基点，并导致实际 GDP 下降至

0.7 %

。

7 | Increases in Bank Capital Requirements or Structural Breaks?
7 | 银行资本要求的增加或结构性断裂？

The validity of our discussion so far hinges on the implicit assumption that the estimated relationships in the VAR have maintained stable throughout the estimation period. If that was not the case, then the discrepancies between the observed developments in the Tier 1 ratio and the conditional forecast would not only reflect the size of the policy actions during the three periods we analyze. For example, Aastveit et al. (2017) document that instability in the estimated relationships do explain a large part of the discrepancies between the main US macroeconomic variables (the unemployment rate, in particular) and their conditional out-of-sample forecasts, in the aftermath of the Global Financial Crisis. In our case, an example may be related to the role of the sovereign spread, which was virtually flat up until the breakout of the sovereign debt crisis and then suddenly increased in 2011. In this context, it is very likely that the relationship between sovereign yields, lending conditions and bank capital has undergone a dramatic time-variation.

^{42}

我们迄今讨论的有效性依赖于一个隐含假设，即 VAR 中估计的关系在整个估计期间保持稳定。如果情况并非如此，那么 Tier 1 比率的观察发展与条件预测之间的差异不仅仅反映了我们分析的三个时期内政策行动的规模。例如，Aastveit 等人（2017）记录了估计关系的不稳定性确实解释了主要美国宏观经济变量（特别是失业率）与其条件外样本预测之间的很大一部分差异，这发生在全球金融危机之后。在我们的案例中，一个例子可能与主权利差的作用有关，该利差在主权债务危机爆发之前几乎保持平稳，然后在 2011 年突然增加。在这种情况下，主权收益率、贷款条件和银行资本之间的关系很可能经历了剧烈的时间变化。

The choice of a relatively short time horizon for the conditional forecasts (2 years) limits these potential concerns. However, the financial and the sovereign debt crises may have induced changes in the relationship between the business cycle, the financial markets, and banking activity quick enough to be visible at short time horizons. Thus, we check the robustness of the main results obtained in the baseline model by relaxing the assumption of fixed-coefficients, for each of the counterfactual described in Section 5. First, we estimate a BVAR model that allows for time-varying coefficients and time-varying volatility (Koop and Korobilis 2013). Second, we use “in-sample” forecasts, so that the underlying relationships in the VAR are estimated using the information up to the end of each forecast window (Aastveit et al. 2017). While the approach of Koop and Korobilis (2013) provides a fully-fledged out-of-sample model, the “in-sample” analysis has more the flavor of a-furthergoodness of fit of the model.
选择相对较短的时间范围进行条件预测（2 年）限制了这些潜在的担忧。然而，金融危机和主权债务危机可能已经引发了商业周期、金融市场和银行活动之间关系的变化，这种变化足够快，以至于在短时间范围内可见。因此，我们通过放宽固定系数的假设，检查基准模型中获得的主要结果的稳健性，针对第 5 节中描述的每个反事实。首先，我们估计一个允许时间变化系数和时间变化波动性的 BVAR 模型（Koop 和 Korobilis 2013）。其次，我们使用“样本内”预测，以便在每个预测窗口结束时使用信息来估计 VAR 中的基本关系（Aastveit 等，2017）。虽然 Koop 和 Korobilis（2013）的方法提供了一个完整的样本外模型，但“样本内”分析更像是对模型的进一步拟合优度检验。

7.1 | Time-Varying Coefficients and Volatility
7.1 | 时间变动系数与波动性

We first consider a BVAR model with time-varying (TV) parameters and volatilities. As already mentioned in Section 4.2, we use the approach proposed by Koop and Korobilis (2013), which introduces shortcuts to alleviate computational constraints in large macroeconomic models, making them more tractable. The estimated size of the policy actions and the associated evolution of the variables of interest over the three episodes are presented in Figures 5-10. Results for the three windows are also summarized in Table 2.
我们首先考虑一个具有时间变化（TV）参数和波动性的 BVAR 模型。如第 4.2 节中所提到的，我们采用 Koop 和 Korobilis（2013）提出的方法，该方法引入了捷径以缓解大型宏观经济模型中的计算限制，使其更易处理。政策行动的估计规模及相关变量在三个阶段的演变呈现在图 5-10 中。三个窗口的结果也在表 2 中进行了总结。

In more detail, Figures 5-7 compare the results between fixed and time-varying coefficients models relying on the counterfactual A described in Section 5 (i.e., the counterfactual Tier 1 ratio based on the BVAR unconditional forecast). In each figure, the blue solid line refers to the estimates based on the baseline (fixed-coefficients) BVAR model, together with percentiles of the posterior distribution (shaded gray area); these are compared with the estimates obtained with the timevarying coefficient model and the correspondent percentiles of the posterior (respectively, the cyan solid, dashed, and dotted lines).
更详细地说，图 5-7 比较了基于第 5 节中描述的反事实 A（即基于 BVAR 无条件预测的反事实一级资本充足率）的固定和时间变动系数模型之间的结果。在每个图中，蓝色实线表示基于基线（固定系数）BVAR 模型的估计值，以及后验分布的百分位数（阴影灰色区域）；这些与时间变动系数模型获得的估计值及其相应的后验百分位数（分别为青色实线、虚线和点线）进行比较。

The analysis qualitatively confirms all the results obtained with the fixed-coefficients model. From a quantitative point of view, the time-varying model leads to larger effects on GDP and loans to HH in the EBA episode and-to a lesser extent-on margins to NFCs in the SSM episode. These differences tend to reflect time variation in the volatility of innovations in some of the estimated equations. Indeed, Figure 11 shows that the model captures large innovations in several equations over the last decade. In particular, the volatility of the real GDP and the short-term rate equations spiked with the breakout of the Global Financial Crisis; instead, large innovations in the Tier 1 equation were observed since late 2013, coinciding with the launch of the Comprehensive Assessment and the counterfactual Tier 1 ratio based on the BVAR conditional forecast obtained using vintages of the MPE). In each figure, the magenta solid line refers to the estimates based on the baseline (fixed-coefficients) BVAR model, together with percentiles of the posterior distribution (shaded gray area); as in the previous figure, the cyan (solid, dashed and dotted) lines show the results of the time-varying coefficient model. The results broadly confirm the picture that emerges when comparing fixed to time-varying coefficient IRFs based on unconditional forecast. One exception is the response of margins to HHs, where the increase is smaller than the one in the fixed-coefficient framework.
分析定性地确认了固定系数模型所获得的所有结果。从定量的角度来看，时变模型在 EBA 事件中对 GDP 和家庭贷款产生了更大的影响，而在 SSM 事件中对非金融公司（NFC）的利润率影响则相对较小。这些差异往往反映了某些估计方程中创新波动性的时间变化。实际上，图 11 显示该模型在过去十年中捕捉到了多个方程中的大创新。特别是，实际 GDP 和短期利率方程的波动性在全球金融危机爆发时激增；相反，自 2013 年底以来，一级资本方程中观察到了大创新，这与全面评估的启动以及基于使用 MPE 的复古数据获得的 BVAR 条件预测的反事实一级资本比率相吻合。在每个图中，品红色实线表示基于基线（固定系数）BVAR 模型的估计，以及后验分布的百分位数（阴影灰色区域）；与前一个图相同，青色（实线、虚线和点线）显示了时间变动系数模型的结果。结果大致确认了在比较基于无条件预测的固定与时间变动系数 IRF 时出现的图景。一个例外是对 HHs 的边际反应，其中的增加小于固定系数框架中的增加。

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of unconditional forecasts, obtained by a simulation of 5,000 draws; the straight cyan line is the policy variable out-of-sample TV-VAR unconditional forecast while the dashed (dotted) cyan lines represent

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts, obtained by a simulation of 5,000 draws. Estimation sample is 1993:Q1-2009:Q1. Second-Eighth panel: the blue line is the difference between the policy and no policy scenario defined in terms of the unconditional forecast and the dark (light) grey shaded area represents

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of this difference; the straight cyan line is the difference (either percentage or basis points) between the policy and no policy scenario defined in terms of the TV-VAR unconditional forecast and the dashed (dotted) cyan lines represent

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts. Forecasting sample is 2009:Q2-2010:Q4.
备注：第一面板：黑线是政策变量（一级资本充足率）的实际数据；蓝线是其样本外 VAR 无条件预测，深灰（浅灰）阴影区域是

16 - 84 % (5 - 95 %)

百分位数的无条件预测经验后验分布，通过 5000 次抽样模拟获得；直线青色线是政策变量样本外 TV-VAR 无条件预测，而虚线（点线）青色线表示

16 - 84 % (5 - 95 %)

百分位数的 TV-VAR 无条件预测经验后验分布，通过 5000 次抽样模拟获得。估计样本为 1993 年第一季度至 2009 年第一季度。第二至第八面板：蓝线是政策与无政策情景之间的差异，以无条件预测为定义，深灰色（浅灰色）阴影区域代表

16 - 84 %

（

5 - 95 %

）的经验后验分布的百分位数；直线青色线是政策与无政策情景之间的差异（以百分比或基点表示），以 TV-VAR 无条件预测为定义，虚线（点线）青色线代表

16 - 84 %

（

5 - 95 %

）的 TV-VAR 无条件预测的经验后验分布的百分位数。预测样本为 2009 年第二季度至 2010 年第四季度。

FIGURE 5 | Fixed vs. time-varying coefficients models: The effects of increasing bank capital requirements, Basel III episode; Counterfactual A.
图 5 | 固定与时间变化系数模型：增加银行资本要求的影响，巴塞尔协议 III 事件；反事实 A。

SSM; significant innovations in the consumer price equation appeared after 2008, showing only mild changes in the recent low-inflation period.
SSM; 在 2008 年后，消费者价格方程出现了显著的创新，在最近的低通胀时期仅显示出轻微的变化。

Similarly, Figures 8-10 compare the results between fixedcoefficients and time-varying coefficients models referring to the counterfactual B described in Section 5 (i.e., the
同样，图 8-10 比较了固定系数模型和时间变动系数模型的结果，参考了第 5 节中描述的反事实 B（即，

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts, obtained by a simulation of 5,000 draws. Estimation sample is 1993:Q1-2010:Q4. Second-Eighth panel: the blue line is the difference between the policy and no policy scenario defined in terms of the unconditional forecast and the dark (light) grey shaded area represents

16 - 84 %

(

5 - 95 %

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts. Forecasting sample is 2011:Q1 - 2012:Q4.
备注：第一面板：黑线是政策变量（一级资本充足率）的实际数据；蓝线是其样本外 VAR 无条件预测，深（浅）灰色阴影区域是

16 - 84 % (5 - 95 %)

百分位数的无条件预测经验后验分布，通过 5000 次抽样模拟获得；直线青色线是政策变量样本外 TV-VAR 无条件预测，而虚线（点线）青色线代表

16 - 84 % (5 - 95 %)

百分位数的 TV-VAR 无条件预测经验后验分布，通过 5000 次抽样模拟获得。估计样本为 1993 年第一季度至 2010 年第四季度。第二至第八面板：蓝线是政策与无政策情景之间的差异，以无条件预测为定义，深灰色（浅灰色）阴影区域代表

16 - 84 %

（

5 - 95 %

16 - 84 %

（

5 - 95 %

）的 TV-VAR 无条件预测的经验后验分布的百分位数。预测样本为 2011 年第一季度至 2012 年第四季度。

FIGURE 6 | Fixed vs. time-varying coefficients models: The effects of increasing bank capital requirements, EBA episode; Counterfactual A.
图 6 | 固定与时间变动系数模型：增加银行资本要求的影响，EBA 事件；反事实 A。

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts, obtained by a simulation of 5,000 draws. Estimation sample is 1993:Q1-2013:Q4. Second-Eighth panel: the blue line is the difference between the policy and no policy scenario defined in terms of the unconditional forecast and the dark (light) grey shaded area represents

16 - 84 %

(

5 - 95 %

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts. Forecasting sample is 2014:Q1-2015:Q4.
备注：第一面板：黑线是政策变量（一级资本比率）的实际数据；蓝线是其样本外 VAR 无条件预测，深灰（浅灰）阴影区域是

16 - 84 % (5 - 95 %)

百分位数的无条件预测经验后验分布，通过 5000 次抽样模拟获得；直线青色线是政策变量样本外 TV-VAR 无条件预测，而虚线（点线）青色线代表

16 - 84 % (5 - 95 %)

百分位数的 TV-VAR 无条件预测经验后验分布，通过 5000 次抽样模拟获得。估计样本为 1993 年第一季度至 2013 年第四季度。第二至第八面板：蓝线是政策与无政策情景之间的差异，以无条件预测为定义，深灰色（浅灰色）阴影区域代表

16 - 84 %

（

5 - 95 %

16 - 84 %

（

5 - 95 %

）的 TV-VAR 无条件预测的经验后验分布的百分位数。预测样本为 2014 年第一季度至 2015 年第四季度。

FIGURE 7 | Fixed vs. time-varying coefficients models: The effects of increasing bank capital requirements, SSM episode; Counterfactual A.
图 7 | 固定与时间变化系数模型：增加银行资本要求的影响，SSM 事件；反事实 A。

7.2 | "In-Sample" Counterfactuals
7.2 | “样本内”反事实

Our second approach to address potential structural breaks in the estimated relationships consists in performing an “in-sample” evaluation exercise, in which we estimate the model up to the end of each forecast horizon and use the estimated coefficients to generate the conditional forecasts over the same horizon. This procedure allows the estimated coefficients to embody potential changes in the relationships between the banking and the macroeconomic variables beyond what could be done using of econometric models that allow for time variation.
我们第二种解决估计关系中潜在结构性断裂的方法是进行“样本内”评估练习，在该练习中，我们估计模型直到每个预测区间的结束，并使用估计的系数在相同的区间内生成条件预测。该程序允许估计的系数体现银行与宏观经济变量之间关系的潜在变化，超出使用允许时间变化的计量经济模型所能做到的范围。

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

16 - 84 %

(

5 - 95 %

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts. Forecasting sample is 2014:Q1 - 2015:Q4.
备注：第一面板：黑线是政策变量（一级资本比率）的实际数据；蓝线是其样本外 VAR 无条件预测，深灰（浅灰）阴影区域是

16 - 84 % (5 - 95 %)

百分位数的无条件预测经验后验分布，通过 5000 次抽样模拟获得；直线青色线是政策变量样本外 TV-VAR 无条件预测，而虚线（点线）青色线代表

16 - 84 % (5 - 95 %)

16 - 84 %

（

5 - 95 %

16 - 84 %

（

5 - 95 %

）的 TV-VAR 无条件预测的经验后验分布的百分位数。预测样本为 2014 年第一季度至 2015 年第四季度。

FIGURE 8 I Fixed vs. time-varying coefficients models: The effects of increasing bank capital requirements, Basel III episode; Counterfactual B.
图 8 I 固定与时间变动系数模型：增加银行资本要求的影响，巴塞尔协议 III 事件；反事实 B。

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

16 - 84 %

(

5 - 95 %

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts. Forecasting sample is 2014:Q1 - 2015:Q4.
备注：第一面板：黑线是政策变量（一级资本比率）的实际数据；蓝线是其样本外 VAR 无条件预测，深灰（浅灰）阴影区域是

16 - 84 % (5 - 95 %)

百分位数的无条件预测经验后验分布，通过 5000 次抽样模拟获得；直线青色线是政策变量样本外 TV-VAR 无条件预测，而虚线（点线）青色线代表

16 - 84 % (5 - 95 %)

16 - 84 %

（

5 - 95 %

16 - 84 %

（

5 - 95 %

）的 TV-VAR 无条件预测的经验后验分布的百分位数。预测样本为 2014 年第一季度至 2015 年第四季度。

FIGURE 9 | Fixed vs. time-varying coefficients models: The effects of increasing bank capital requirements, EBA episode; Counterfactual B.
图 9 | 固定与时间变动系数模型：增加银行资本要求的影响，EBA 事件；反事实 B。

Figure 12 reports the estimation results, for the Tier 1 ratio and the size of the policy actions, based on in-sample forecasts produced both with the fixed-coefficient and the time-varying coefficient models.
图 12 报告了基于固定系数和时间变动系数模型产生的样本内预测的一级资本比率和政策行动规模的估计结果。

This exercise leads to very conservative estimates of the size of the policy actions: The increase in the capital requirement after two years is only about half than that in the baseline model in the EBA episode and about one-fifth in the SSM/CA episode.
这一练习导致对政策行动规模的非常保守的估计：在 EBA 事件中，两年后资本要求的增加仅为基线模型的一半，而在 SSM/CA 事件中约为五分之一。

Notes: First panel: the black line is the actual data on the policy variable (Tier 1 ratio); the blue line is its out-of-sample VAR unconditional forecast and the dark (light) grey shaded area is

16 - 84 % (5 - 95 %)

16 - 84 % (5 - 95 %)

16 - 84 %

(

5 - 95 %

16 - 84 %

(

5 - 95 %

) percentiles of the empirical posterior distribution of TV-VAR unconditional forecasts. Forecasting sample is 2014:Q1 - 2015:Q4.
备注：第一面板：黑线是政策变量（一级资本比率）的实际数据；蓝线是其样本外 VAR 无条件预测，深灰（浅灰）阴影区域是

16 - 84 % (5 - 95 %)

百分位数的无条件预测经验后验分布，通过 5000 次抽样模拟获得；直线青色线是政策变量样本外 TV-VAR 无条件预测，而虚线（点线）青色线代表

16 - 84 % (5 - 95 %)

16 - 84 %

（

5 - 95 %

16 - 84 %

（

5 - 95 %

）的 TV-VAR 无条件预测的经验后验分布的百分位数。预测样本为 2014 年第一季度至 2015 年第四季度。

FIGURE 10 | Fixed vs. time-varying coefficients models: The effects of increasing bank capital requirements, SSM episode; Counterfactual B.
图 10 | 固定与时间变动系数模型：增加银行资本要求的影响，SSM 事件；反事实 B。
TABLE 2 | Cross-episodes effects of bank capital requirements increases in the time-varying coefficients BVAR.
表 2 | 银行资本要求增加的跨期效应在时间变动系数 BVAR 中。

Episode (2-year window)集 (2 年窗口)

Horizon地平线

政策行动的预计规模（百分点）一级资本充足率

Estimated size of the policy actions (pp)

Tier 1 ratio

Estimated impact after

H

periods
预计在

H

个周期后的影响

Macro variables宏观变量

Loan volume贷款金额

Loan margin (bps)贷款利差（基点）

Real GDP实际国内生产总值

HICP

NFCs

HHs

NFCs

HHs

Unconditional无条件

EBA

0.501

0.000

18.899

24.906

1.551

-0.30

0.001

-0.014

-0.021

77.100

95.512

2.533

-0.031

-0.008

-0.049

-0.027

94.812

116.364

SSM/CA

0.157

0.000

18.513

32.313

0.882

-0.020

0.002

-0.002

-0.009

62.462

22.345

1.746

-0.025

0.001

-0.017

-0.040

64.642

67.200

Average平均

0.329

0.000

18.706

28.610

1.217

-0.025

0.001

-0.008

-0.015

69.781

58.928

2.140

-0.028

-0.004

-0.033

-0.032

79.727

91.782

Conditional条件

EBA

0.502

0.000

20.954

23.313

1.565

-0.030

-0.001

-0.015

-0.022

78.707

93.420

2.527

-0.029

-0.008

-0.051

-0.028

96.073

117.714

SSM/CA

0.263

0.000

15.402

14.380

1.101

-0.016

0.002

-0.005

-0.015

60.581

56.416

1.881

-0.030

0.006

-0.041

-0.047

103.805

96.494

Average平均

0.383

0.000

15.402

14.380

1.333

-0.023

0.000

-0.010

-0.018

69.644

74.918

2.204

-0.030

-0.001

-0.046

-0.038

99.939

107.104

B. Macroeconomic and credit variables
B. 宏观经济和信用变量

Abstract摘要

Notes: The upper panel plots the estimated time-varying volatility of the Tier 1 ratio equation in the BVAR model (median of the posterior distribution). The lower panel presents the same statistics for real GDP, the HICP, the short-term interest rate, the long-term interest rate, loans and margins to non-financial corporations, loans and margins to households. Estimation sample is 1993:Q1 - 2015:Q4.
备注：上面的面板绘制了 BVAR 模型中一级资本比率方程的估计时间变化波动性（后验分布的中位数）。下面的面板呈现了实际 GDP、HICP、短期利率、长期利率、对非金融公司的贷款和利差、对家庭的贷款和利差的相同统计数据。估计样本为 1993 年第一季度至 2015 年第四季度。

FIGURE 11 | Time-varying volatility.
图 11 | 时间变化的波动性。

Nonetheless, the observed ratio remains outside the bands even under this very conservative approach, a fact which we interpret as strongly confirming the validity of our results. Consequently, the estimated impact on GDP, lending volumes, and lending rates based on this more conservative approach have the same signs of those in the baseline and in the time-varying approaches, though being somewhat lower.

^{43}

尽管如此，观察到的比率在这种非常保守的方法下仍然超出范围，我们将这一事实解读为强烈确认我们结果的有效性。因此，基于这种更保守的方法对 GDP、贷款量和贷款利率的估计影响与基线和时间变化方法中的符号相同，尽管略低。

^{43}

8 | Additional Analyses and Robustness
8 | 附加分析与稳健性

In this section, we discuss a number of additional exercises and robustness checks, which further strengthen our findings. A few additional less important checks on which we comment in the paper are reported in the Supporting Information Appendix, such as including the shadow rate to better capture unconventional monetary policy.
在本节中，我们讨论了一些额外的练习和稳健性检验，这进一步加强了我们的发现。我们在论文中评论的一些额外的较不重要的检验报告在支持信息附录中，例如包括影子利率以更好地捕捉非常规货币政策。

8.1 | Evidence From the Euro Area BLS
8.1 | 来自欧元区 BLS 的证据

The Euro Area BLS, conducted quarterly by the Eurosystem, regularly collects information from a large sample of Euro-area banks on credit demand and supply conditions to the private sector. In addition, it regularly includes additional (ad hoc) questions on various topics, including banks’ balance sheet situations. In particular, since 2011, the survey has included-with a 6 -month frequency-one question asking the banks whether new or expected regulatory and/or supervisory actions had an impact on their balance sheets and, in particular, on their capital position; moreover, the question asks to distinctively indicate the impact on equity and RWAs.

^{44}

欧元区银行贷款调查（BLS）由欧洲系统每季度进行，定期收集来自大量欧元区银行的关于对私营部门的信贷需求和供应状况的信息。此外，它还定期包括关于各种主题的额外（临时）问题，包括银行的资产负债表状况。特别是，自 2011 年以来，该调查以 6 个月的频率包括一个问题，询问银行新的或预期的监管和/或监督措施是否对其资产负债表产生了影响，特别是对其资本状况的影响；此外，该问题要求明确指出对股本和风险加权资产（RWAs）的影响。

^{44}

Figure 13 shows that about

30 %

of the banks participating to the BLS had to increase their capital in the second half of 2011 and 2012 as a consequence of requests from the regulator/supervisor; moreover, about

20 %

had to reduce RWAs. In the SSM/CA episode, we also observe a significant increase in the share of banks reporting that these policy actions had an impact on their capital position, especially via additional equity issuance. While the chart is drawn for the entire sample of Euro area banks, there is no reason to believe that replies for the subsample of Italian banks-which are are not publicly available-are substantially different. This evidence strongly supports the notion that increases in banks’ capital ratios during the EBA and the SSM/CA episode were driven by regulatory/supervisory actions.
图 13 显示，大约

30 %

的参与 BLS 的银行在 2011 年下半年和 2012 年因监管机构/监督机构的要求不得不增加其资本；此外，大约

20 %

不得不减少风险加权资产（RWAs）。在 SSM/CA 事件中，我们还观察到报告这些政策措施对其资本状况产生影响的银行比例显著增加，特别是通过额外的股权发行。虽然该图是针对整个欧元区银行样本绘制的，但没有理由相信意大利银行的子样本的回复——这些回复并未公开——会有实质性差异。这一证据强烈支持了在 EBA 和 SSM/CA 事件期间，银行资本比率的增加是由监管/监督行动驱动的观点。

Notes: See Figure 2-4. The major difference is that estimation sample is 1993:Q1-2015:Q4 for Basel III, EBA and the SSM, whereas the forecasting windows are, respectively, 2009:Q2-2011:Q1, 2011:Q1-2012:Q4 and 2014:Q1-2015:Q4.
注：见图 2-4。主要区别在于，巴塞尔协议 III、EBA 和 SSM 的估计样本为 1993 年第一季度至 2015 年第四季度，而预测窗口分别为 2009 年第二季度至 2011 年第一季度、2011 年第一季度至 2012 年第四季度和 2014 年第一季度至 2015 年第四季度。

FIGURE 12 | “In-sample” forecasts of the Tier 1 ratio.
图 12 | “样本内”预测的一级资本比率。

8.2 | Forecasts of Main Macroeconomic and Credit Variables Under the Policy Scenario
8.2 | 政策情景下主要宏观经济和信贷变量的预测

One useful exercise to corroborate the findings of the scenario analysis is to check that the conditional forecasts of GDP, HICP, lending, and lending rates in the policy scenario (i.e., conditional on the actual path of the Tier 1 ratio) are in line with their actual developments. If that were not the case, one may be concerned that other major events drove the dynamics of the Tier 1 ratio rather to tighter supervisory/regulatory pressure.
一个有用的练习来证实情景分析的发现是检查政策情景下（即，基于实际的一级资本比率路径）的 GDP、HICP、贷款和贷款利率的条件预测是否与它们的实际发展一致。如果不是这样，人们可能会担心其他重大事件推动了一级资本比率的动态，而不是更严格的监督/监管压力。

Figure G1 (reported in the Supporting Information Appendix G) plots actual developments of our main macro-financial variables of interest against their conditional forecasts in the policy scenario, in the EBA and the SSM/CA episodes. In both cases, actual developments of the main variables fall within the credibility interval of their forecast under the policy scenario. The only two exceptions are the HICP level and the margins applied to loans to NFCs: As for the former, other studies find it very difficult to forecast price dynamics after the Global Financial Crisis and, especially, since

2012^{45}

; as for margins, the reduction observed in the 2014-2015 is likely to reflect ECB monetary policy easing. All in all, this exercise broadly corroborates our interpretation that the dynamics of the Tier 1 ratio during the two episodes can be safely attributed to increased pressure from the regulators/supervisors.
图 G1（在支持信息附录 G 中报告）绘制了我们主要宏观金融变量的实际发展与政策情景下的条件预测之间的关系，分别在 EBA 和 SSM/CA 事件中。在这两种情况下，主要变量的实际发展均落在其政策情景预测的可信区间内。唯一的两个例外是 HICP 水平和适用于非金融公司贷款的利差：就前者而言，其他研究发现自全球金融危机以来，尤其是自

2012^{45}

以来，预测价格动态非常困难；至于利差，2014-2015 年观察到的减少可能反映了欧洲央行的货币政策宽松。总的来说，这项工作大致证实了我们的解释，即在这两个事件中，一级资本比率的动态可以安全地归因于来自监管机构/监督者的压力增加。

8.3 | Including the US Tier 1 Ratio as Global Regulation Factor
8.3 | 包括美国一级资本充足率作为全球监管因素

A possible concern about our findings is given by the possibility of omitted variables related to regulation. Are we really capturing the regulatory/supervisory policy actions and their estimated impact on the Italian economy? Or are we instead confounding these effects because of some omitted variables linked to the global process of regulation triggered by the Global Financial Crisis? To answer this question, we add the US Tier 1 ratio to the baseline BVAR model and we replicate our scenario analysis. The advantage of this choice is twofold. First, the US Tier 1 ratio can be interpreted as a proxy of a global regulatory factor. In practice, the inclusion of this variable in the BVAR cleans the dynamics by a common regulation component, leaving more confidence on the fact that we are indeed measuring Italy’s idiosyncratic developments in the Tier 1 ratio. Second, the dynamics of the US Tier 1 ratio could account for bank capital evolution unrelated to the EBA and SSM policy initiatives, therefore being useful as a “control” variable in the counterfactual setup.
我们发现的一个可能问题是与监管相关的遗漏变量的可能性。我们是否真的捕捉到了监管/监督政策行动及其对意大利经济的估计影响？还是说我们因为一些与全球金融危机引发的全球监管过程相关的遗漏变量而混淆了这些影响？为了解答这个问题，我们将美国的一级资本比率添加到基线 BVAR 模型中，并复制我们的情景分析。这一选择的优势有两个方面。首先，美国的一级资本比率可以被解释为全球监管因素的代理。在实践中，将该变量纳入 BVAR 模型可以通过一个共同的监管成分清理动态，从而更有信心地认为我们确实在测量意大利一级资本比率的特有发展。其次，美国一级资本比率的动态可能与 EBA 和 SSM 政策倡议无关的银行资本演变相关，因此在反事实设置中作为“控制”变量是有用的。

Notes: Net percentages; half-yearly frequency answers. For “risk-weighted assets” and “capital”, the net percentages are defined as the difference between the sum of the percentages of answers “increased considerably” and “increased somewhat” and the sum of the percentages of answers “decreased somewhat” and “decreased considerably”. The same methodology applies to “average loans”, “riskier loans”, “retained earnings” and “capital issuance”. For further details, see the BLS website at https://www.ecb.europa.eu/stats/ecb_ surveys/bank_lending_survey/html/index.en.html.
注意：净百分比；半年频率回答。对于“风险加权资产”和“资本”，净百分比定义为“显著增加”和“略微增加”回答的百分比总和与“略微减少”和“显著减少”回答的百分比总和之间的差值。同样的方法适用于“平均贷款”、“风险较高的贷款”、“留存收益”和“资本发行”。有关更多详细信息，请参见 BLS 网站 https://www.ecb.europa.eu/stats/ecb_surveys/bank_lending_survey/html/index.en.html。

FIGURE 13 | Euro area bank lending survey: The impact of regulatory and supervisory actions on bank’s capital ratio.
图 13 | 欧元区银行贷款调查：监管和监督措施对银行资本比率的影响。

We add the US Tier 1 ratio to the baseline BVAR,

^{46}

thus increasing the dimension to

N = 17

variables. After estimating the extended model, we modify the way in which we compute the counterfactual paths of Italy’s Tier 1 ratio by adding the US Tier 1 ratio as a conditioning variable. More explicitly, we now obtain the two counterfactuals for the Italian Tier 1 ratio as follows: (i) the forecast conditional (only) on the actual US Tier 1 ratio and (ii) the forecast conditional on the US Tier 1 ratio and the expected values of the main drivers of bank capital in real time (i.e., the same variables as in the baseline). Unfortunately, we were unable to find real-time vintages for the US Tier 1 ratio: We use the actual values, that can reasonably be considered exogenous with respect to developments in the Italian Tier 1 ratio. In practice, the counterfactual Italy’s Tier 1 ratio is now also influenced by the evolution of US capital ratio, a feature which may help to remove our estimated size of the policy actions from a global/common component of banks’ Tier 1 ratios and obtain a sharper estimate of the idiosyncratic Italian developments. The results are presented in Figures D2 and D3 in the Supporting Information Appendix D and broadly confirm all the main findings of the baseline 16 variable-model. The only exception is a somewhat larger uncertainty on the IRFs of margins to NFC during the EBA episode. In this regard, it should be noted that the US Tier 1 ratio is only available since 1996:Q1, which is somewhat unfortunate, as we are forced to drop observations 1993:Q11995:Q4, a period in which loans to NFC growth experienced a slowdown (from over

to about

) after the contraction in economic activity associated with the EMS Crisis of 1992-1993. It has to be kept in mind that this is the stronger recession experimented by the Italian economy before the Global Financial Crisis: Dropping these observations hence leads to some loss of information in the business cycle analysis. All in all, the inclusion of the US Tier 1 ratio provides simple and effective evidence on the robustness of our conclusions presented in Section 6.
我们将美国的一级资本充足率添加到基线 BVAR 中，

^{46}

从而将维度增加到

N = 17

个变量。在估计扩展模型后，我们修改了计算意大利一级资本充足率反事实路径的方法，通过将美国一级资本充足率作为条件变量。更明确地说，我们现在获得了意大利一级资本充足率的两个反事实，如下所示：(i) 仅基于实际美国一级资本充足率的预测，以及 (ii) 基于美国一级资本充足率和实时主要银行资本驱动因素的预期值的预测（即，与基线中的相同变量）。不幸的是，我们无法找到美国一级资本充足率的实时数据：我们使用实际值，这些值可以合理地被视为与意大利一级资本充足率的发展无关。实际上，意大利的反事实一级资本充足率现在也受到美国资本充足率演变的影响，这一特征可能有助于将我们估计的政策行动规模从银行一级资本充足率的全球/共同组成部分中剔除，并获得对意大利特有发展的更精确估计。结果在支持信息附录 D 的图 D2 和 D3 中呈现，并广泛确认了基线 16 变量模型的所有主要发现。唯一的例外是在 EBA 事件期间，NFC 的边际影响的脉冲响应函数（IRFs）存在较大的不确定性。在这方面，需要注意的是，美国的一级资本比率自 1996 年第一季度以来才有数据，这有些不幸，因为我们被迫放弃 1993 年第一季度至 1995 年第四季度的观察数据，这一时期 NFC 贷款增长经历了放缓（从超过

降至约

），这是由于 1992-1993 年 EMS 危机导致的经济活动收缩。必须记住，这是意大利经济在全球金融危机之前经历的最严重的衰退。因此，放弃这些观察数据会导致商业周期分析中一些信息的丧失。总的来说，纳入美国的一级资本比率为我们在第 6 节中提出的结论的稳健性提供了简单有效的证据。

8.4 | A "Placebo" Test
8.4 | 一项“安慰剂”测试

We have shown that our model is able to correctly anticipate the dynamics of Tier 1 ratio in an out-of-sample setting for the Basel III case. Nevertheless, a natural question that arises after looking at our findings for the EBA and the SSM is whether an empirical exercise conducted outside of the three periods of interest for the regulatory policy should produce predicted capital ratios whose credible interval covers the actual capital ratio. Put it differently, one may be worried that there is some systematic deviation between the actual and the forecast path of the Tier 1 ratio. To investigate this aspect, the 2012-2013 period seems appropriate, as 2012 lies at the end of the second exercise, while 2013 lies outside the policy action periods. Running our empirical analysis over this period would hence represent a sort of “placebo” test, reinforcing the genuine nature of the findings provided in Section 6.
我们已经证明我们的模型能够在巴塞尔 III 案例的样本外环境中正确预测一级资本比率的动态。然而，在查看我们对 EBA 和 SSM 的发现后，自然会产生一个问题，即在监管政策的三个关注时期之外进行的实证研究是否应该产生其可信区间覆盖实际资本比率的预测资本比率。换句话说，人们可能会担心实际与预测的一级资本比率路径之间存在某种系统性偏差。为了调查这一方面，2012-2013 年期间似乎是合适的，因为 2012 年位于第二次实证研究的结束，而 2013 年则位于政策行动时期之外。因此，在这一时期进行我们的实证分析将代表一种“安慰剂”测试，进一步强化第 6 节中提供的发现的真实性。

To further enhance the soundness of our empirical strategy and findings, we therefore conduct this “placebo” test as follows. We estimate the model until 2012:Q1 and then evaluate the IRFs in a window of 8 quarters. The results are shown in Figure F1 in the Supporting Information Appendix F. Two remarks are in order. First, when using the unconditional forecast of the Tier 1 ratio to obtain the evolution of

{\hat{k}}_{t}

-that is, the counterfactual Tier 1 ratio-the credibility interval covers the actual capital ratio: In fact, the median of the posterior distribution of the unconditional forecasts is very close to, although somewhat higher than, the actual developments. This implies a wide uncertainty in the estimated effects of the policy actions, that is, the credibility interval of variables of interest is wide around the zero. Second, when using the conditional forecast based on the vintages of the Eurosystem Macroeconomic Projection Database, the counterfactual Tier 1 ratio overestimates the actual dynamics, mainly due to overly favorable patterns of economic activity in the Eurosystem projections, that is, positive forecast errors. Consistently, the IRFs show expansionary effects on economic activity and lending supply. Overall, the evidence provided by this exercise confirms the absence of systematic (under) deviations of actual bank capital from its expected path. This corroborates and strengthens our empirical findings across the three major episodes of rising Tier 1 ratio (Basel III, EBA, and SSM).
为了进一步增强我们实证策略和发现的可靠性，我们因此进行如下“安慰剂”测试。我们估计模型直到 2012 年第一季度，然后在 8 个季度的窗口中评估脉冲响应函数（IRFs）。结果显示在支持信息附录 F 的图 F1 中。有两点需要说明。首先，当使用无条件预测的一级资本比率来获得

{\hat{k}}_{t}

的演变——即反事实的一级资本比率——时，可信区间覆盖了实际资本比率：实际上，无条件预测的后验分布的中位数非常接近，尽管略高于实际发展。这意味着政策行动的估计效果存在较大的不确定性，即感兴趣变量的可信区间围绕零较宽。其次，当使用基于欧洲系统宏观经济预测数据库的不同版本的条件预测时，反事实的一级资本比率高估了实际动态，主要是由于欧洲系统预测中经济活动的过于有利模式，即正的预测误差。一致地，脉冲响应函数显示出对经济活动和贷款供应的扩张效应。总体而言，这项研究提供的证据确认了实际银行资本与其预期路径之间不存在系统性的（低于）偏差。这证实并加强了我们在三大主要的一级资本比率上升事件（巴塞尔协议 III、欧洲银行管理局和单一监管机制）中的实证发现。

8.5 | Separating Tier 1 Capital and RWAs
8.5 | 分离一级资本和风险加权资产

In our baseline BVAR, we use the Tier 1 ratio as regulatory/supervisory policy variable, consistently with the literature and the policy implementation which is focused on the ratio between the components, bank capital and RWA. However, it would be interesting to try to disentangle the effect of changes in the outstanding Tier 1 capital from those in RWA in the counterfactual simulation. This could shed light on (i) the channel of adjustment of Italian banks and (ii) the channels of impact on the real economy, for example, the crowding-out effect on equity if higher capital ratios came from equity issuance vs. some form of rationing on productive credit if higher ratios came from rebalancing away from productive and risky sectors. Indeed, Figure 1 suggests that the first episode of interest (Basel III) gave rise to an increase in outstanding Tier 1 capital but stopped the fall in risk weighted assets which, ceteris paribus, should have had an expansionary impact. While this is only a graphical analysis and the plateauing of RWA may come from other economic developments, this way of increasing the Tier 1 capital ratio is dramatically different from the next two episodes, where RWA fell while outstanding Tier 1 increased. Further investigation of this should thus be key to understanding the channels at play.
在我们的基准 BVAR 中，我们使用一级资本比率作为监管/监督政策变量，这与文献和政策实施一致，后者集中于银行资本与风险加权资产（RWA）之间的比率。然而，尝试在反事实模拟中区分未偿还的一级资本变化与 RWA 变化的影响将是有趣的。这可以阐明（i）意大利银行的调整渠道和（ii）对实体经济的影响渠道，例如，如果更高的资本比率来自股权发行，则对股权的挤出效应与如果更高的比率来自于从生产性和高风险部门的再平衡，则对生产性信贷的某种形式的配给。实际上，图 1 表明，第一次关注的事件（巴塞尔协议 III）导致未偿还的一级资本增加，但停止了风险加权资产的下降，其他条件不变，这应该产生扩张性影响。虽然这仅仅是一个图形分析，RWA 的平稳可能源于其他经济发展，但这种提高一级资本比率的方式与接下来的两个事件截然不同，在这两个事件中，RWA 下降而未偿还的一级资本增加。因此，进一步调查这一点应该是理解相关渠道的关键。

In order to address this issue, we separate numerator and denominator of the Tier 1 ratio, hence estimating a VAR model including

N = 17

variables. The scenario analysis is conducted accordingly, as we now have two estimated size of the policy actions (see Figures E1 and E2). Again, we stress that the Basel III episode in very well predicted by the VAR. Similar to the model including the Tier 1 ratio, also the one including its separate components, quantifies the relevant size of policy actions in the EBA and SSM episodes. The risk-weighted channel seems to be larger in the EBA episode, whereas capital adjustment was stronger in the SSM episode. The estimated IRFs confirm the results of the baseline model (i.e., the one which includes the Tier 1 ratio). Overall, again, the results delivered by this robustness exercise corroborate the main message of our baseline scenario analysis, that is, the contractionary effects of raising bank capital requirements in the short- to medium-run.
为了应对这个问题，我们将一级资本比率的分子和分母分开，因此估计了一个包含

N = 17

个变量的 VAR 模型。相应地进行情景分析，因为我们现在有两个政策行动的估计规模（见图 E1 和 E2）。我们再次强调，巴塞尔 III 事件在 VAR 中得到了很好的预测。与包含一级资本比率的模型类似，包含其单独组成部分的模型也量化了 EBA 和 SSM 事件中政策行动的相关规模。在 EBA 事件中，风险加权渠道似乎更大，而在 SSM 事件中，资本调整更强。估计的 IRF 确认了基线模型的结果（即包含一级资本比率的模型）。总体而言，这项稳健性检验所提供的结果再次证实了我们基线情景分析的主要信息，即在短期到中期内，提高银行资本要求的收缩效应。

9 | Concluding Remarks
9 | 结论

In this paper, we study the impact of an increase in bank capital requirements on lending supply and economic activity. We estimate a 16 variable-Bayesian VAR model including macroeconomic, financial and banking variables for the Italian economy over the period 1993:Q1-2015:Q4. We offer two main contributions to the literature.
在本文中，我们研究了银行资本要求增加对贷款供应和经济活动的影响。我们估计了一个包含宏观经济、金融和银行变量的 16 变量贝叶斯 VAR 模型，覆盖意大利经济 1993 年第一季度至 2015 年第四季度的时期。我们对文献做出了两个主要贡献。

First, we quantify the short-run impact of some key regulatory/ supervisory initiatives on bank capitalization undertaken in the postcrisis period (Basel III; the EBA stress test; the launch of the SSM and the CA by the ECB) in an economy that has been at the epicenter of the European sovereign debt crisis and whose banking system has been under the deep scrutiny of the supervision authorities. We find that for the EBA and SSM episodes the increase in the Tier 1 ratio required by the regulatory policy is deemed unpredictable by a size of about 2.5 percentage points on average. This translated into (i) a contraction of lending supply between

4 %

and

5 %

of the stock of loans and (ii) an increase in loan margins to firms and households in a range of

80 - 130 bps

(in both cases on whether NFCs or HHs are considered and on the way the counterfactual path of Tier 1 is obtained). In turn, the level of GDP declined by about

2 %

in the EBA episode and by over

3 %

in the SSM/CA episode. In contrast, the increase in Tier 1 ratio observed around the Basel III episode is correctly anticipated by our model, being therefore consistent with historical macroeconomic and banking dynamic correlations. Accordingly, no significant effects on lending supply and economic activity is detected.
首先，我们量化了一些关键监管/监督举措在后危机时期对银行资本化的短期影响（巴塞尔协议 III；欧洲银行管理局压力测试；欧洲央行启动的单一监督机制和共同清算机制），这些举措发生在一个处于欧洲主权债务危机中心的经济体，其银行系统一直受到监管机构的严格审查。我们发现，对于欧洲银行管理局和单一监督机制的事件，监管政策要求的一级资本比率的增加被认为是不可预测的，平均约为 2.5 个百分点。这转化为（i）贷款供应收缩在

4 %

和

5 %

之间，以及（ii）对企业和家庭的贷款利差增加在

80 - 130 bps

的范围内（在这两种情况下，无论是考虑非金融公司还是家庭，以及获得一级资本反事实路径的方式）。反过来，GDP 水平在欧洲银行管理局事件中下降了约

2 %

，在单一监督机制/共同清算机制事件中下降了超过

3 %

。与此相反，我们的模型正确预测了在巴塞尔 III 事件期间观察到的一级资本充足率的增加，因此与历史宏观经济和银行动态相关性一致。因此，未发现对贷款供应和经济活动的显著影响。

As a second contribution, we develop an alternative macroeconomic methodology for measuring the size of regulatory or supervisory requests of capital increases-and the associated effects on the evolution of macroeconomic and banking vari-ables-when information on actual capital requirements is not available. Our methodology relies on scenario analysis in which counterfactual patterns of the Tier 1 ratio are obtained by means of unconditional or conditional forecasting techniques, borrowing from the literature on the effects of unconventional monetary policy (Lenza, Pill, and Reichlin 2010; Giannone et al. 2012; Kapetanios et al. 2012; Dahlhaus, Hess, and Reza 2018; Altavilla, Canova, and Ciccarelli 2020).
作为第二个贡献，我们开发了一种替代的宏观经济方法来衡量监管或监督资本增加请求的规模及其对宏观经济和银行变量演变的相关影响，当实际资本要求的信息不可用时。我们的方法依赖于情景分析，通过无条件或有条件的预测技术获得一级资本充足率的反事实模式，借鉴了关于非常规货币政策影响的文献（Lenza, Pill, and Reichlin 2010; Giannone et al. 2012; Kapetanios et al. 2012; Dahlhaus, Hess, and Reza 2018; Altavilla, Canova, and Ciccarelli 2020）。

The advantage of this approach is not requiring an explicit identification of structural bank capital shocks, like in the standard SVAR literature. Moreover, it provides a quasi-real time assessment, allowing for having different impact of regulatory/ supervisory policy actions on credit and real GDP for different episodes. Our interpretation of the difference between the observed capital ratio and its counterfactual development as a measure of the size of the policy actions during each time window is corroborated by several considerations. First, a compelling narrative during the considered event windows, as the authorities explicitly asked Italian banks to significantly increase their capital levels. Second, the fairly rich characterization of the banking sector is crucial for isolating the impact of regulatory and supervisory interventions on bank capital, as it allows us to consider several potential interactions among developments in the real economy, financial and credit markets, strengthening the out-of-sample forecasting performance of the BVAR. Third, the counterfactual exercises are designed in such a way that neither monetary policy neither other important drivers of slow-moving variables such as prices, loan volumes and-especially-economic activity may be confounding factors with respect to regulation/supervisory policy, hence sharpening the quantification of the effects (Altavilla, Giannone, and Lenza 2016; Rostagno et al. 2019). Finally, we provide a wide array of arguments, additional analyses, and robustness tests aimed at convincing the reader that bank capital is correctly modeled and the scenarios analysis correctly measures the size of the policy actions and their impact on the macroeconomic and credit variables.
这种方法的优势在于不需要像标准 SVAR 文献那样明确识别结构性银行资本冲击。此外，它提供了准实时评估，允许在不同事件中对监管/监督政策措施对信贷和实际 GDP 的不同影响进行分析。我们将观察到的资本比率与其反事实发展之间的差异解读为每个时间窗口内政策措施规模的一个衡量，这一点得到了几个考虑因素的支持。首先，在考虑的事件窗口期间，权威机构明确要求意大利银行显著提高其资本水平，这是一个引人注目的叙述。其次，对银行业的相对丰富的特征描述对于隔离监管和监督干预对银行资本的影响至关重要，因为它使我们能够考虑实际经济、金融和信贷市场发展之间的几种潜在互动，从而增强 BVAR 的样本外预测性能。第三，这些反事实演练的设计方式使得货币政策以及其他重要的慢变量驱动因素，如价格、贷款量，尤其是经济活动，可能不会对监管/监督政策产生混淆因素，从而提高了效果的量化（Altavilla, Giannone, and Lenza 2016; Rostagno et al. 2019）。最后，我们提供了广泛的论据、额外的分析和稳健性测试，旨在说服读者银行资本的建模是正确的，情景分析正确地衡量了政策行动的规模及其对宏观经济和信贷变量的影响。

Since our findings could be affected by changes in the way in which policy actions transmit to the economy or in their volatility, we replicate our analysis allowing for time-varying coefficients and volatility (Koop and Korobilis 2013). Overall, these checks confirm that the effects of the policy actions intended to raise bank capital on lending conditions and real economic activity are sizable and significant also when considering this more conservative definition of the policy actions. Moreover, the estimated impact differs across periods and forecasting models as the result of sudden increases in the stochastic volatility in some estimated equations. This suggests that the evaluation of the effects of policy raising bank capital requirements remains challenging and call for statistical models allowing for time variation in coefficients and volatility. Last, but not the least, as further validation of our approach, our scenario analysis yields findings broadly consistently with those obtained in a companion paper (Conti, Nobili, and Signoretti 2023) which uses a SVAR and narrative sign restrictions techniques à la Antolín-Díaz and Rubio-Ramírez (2018).
由于我们的发现可能受到政策行动传递到经济的方式或其波动性的变化的影响，我们复制了我们的分析，允许时间变化的系数和波动性（Koop 和 Korobilis 2013）。总体而言，这些检查确认了旨在提高银行资本的政策行动对贷款条件和实际经济活动的影响是相当大且显著的，即使在考虑这种更保守的政策行动定义时也是如此。此外，估计的影响在不同的时期和预测模型中有所不同，这是由于某些估计方程中随机波动性的突然增加所导致的。这表明，评估提高银行资本要求的政策效果仍然具有挑战性，并呼吁使用允许系数和波动性时间变化的统计模型。最后但同样重要的是，作为我们方法的进一步验证，我们的情景分析得出的结果与在一篇相关论文中获得的结果大体一致（Conti, Nobili, 和 Signoretti 2023），该论文使用了 SVAR 和叙述性符号限制技术，类似于 Antolín-Díaz 和 Rubio-Ramírez（2018）。

Our results yield a number of important policy implications. First, when increasing bank capital requirements, supervisory authorities should carefully take into account the possible feedback effects between changes in regulatory capital and the macroeconomy: In a low-growth environment, regulatory pressures induce banks to tighten credit supply and reduce real GDP, which, in turn, exert pressure on banks to strengthen their capital position, thus reinforcing the initial negative effects on credit supply and economic activity. Moreover, this negative feedback may affect the transmission of monetary policy, possibly crowding out the effectiveness of expansionary measures, and should be thus taken into account also by central banks. An important aspect to bear in mind is that our focus is on the short-run costs of the reforms in the banking system and our methodology disregards the large long-run benefits of banking regulation and supervision, which improve banks’ resilience to shocks and foster financial stability.
我们的研究结果产生了一些重要的政策含义。首先，在提高银行资本要求时，监管机构应仔细考虑监管资本变化与宏观经济之间可能的反馈效应：在低增长环境中，监管压力促使银行收紧信贷供应并减少实际 GDP，这反过来又对银行施加压力，要求其加强资本状况，从而加剧了对信贷供应和经济活动的初始负面影响。此外，这种负反馈可能影响货币政策的传导，可能挤出扩张性措施的有效性，因此中央银行也应考虑这一点。需要牢记的一个重要方面是，我们的重点是银行系统改革的短期成本，而我们的方法论忽略了银行监管和监督带来的长期巨大好处，这些好处提高了银行对冲击的韧性并促进了金融稳定。

There are several promising avenues for future research. First, assessing state-dependent effects of changes in bank capital re-quirements-as a function of the macro-financial environment, (Lang and Menno 2023)-and the interaction between capital requirements and monetary policy may be particularly relevant. Recently, Mendicino et al. (2020) argued in a calibrated framework that the existence of the zero lower bound may impose severe constraints on the ability of monetary policy to counteract the short-run adverse effects of capital regulation. Our results somewhat address this issue, as we cover a period of unconventional monetary policy, which we explicitly account for by including short- and long-term interest rates in our BVAR and, among the robustness exercises conducted, a shadow interest rate that may be useful for capturing unconventional policies. However, a more direct empirical analysis of the interaction between regulation/supervision and monetary policy interventions - for example, in a nonlinear setting similar to Davidson and Moccero (2024)—would be a significant advance in research. Second, future works could study other countries and benefit from including data on manufacturing, services, and construction to determine which sectors are most affected by increased capital requirements. Our framework can be indeed easily extended, even to a multicountry/sector model. Third, the inclusion of deposit flows and interest rates would allow for a more comprehensive analysis of bank performance and related macroeconomic effects. Finally, extending the sample to cover the Covid-19 pandemic (Altavilla et al. 2023)

^{47}

and the ensuing period of high inflation, when bank’s risk perception severely impacted loan dynamics (Auer and Conti 2024), will broaden our knowledge of the relationship between prudential policy and bank lending.
未来研究有几个有前景的方向。首先，评估银行资本要求变化的状态依赖效应——作为宏观金融环境的一个函数（Lang 和 Menno 2023）——以及资本要求与货币政策之间的互动可能特别相关。最近，Mendicino 等人（2020）在一个校准框架中指出，零下限的存在可能对货币政策抵消资本监管短期不利影响的能力施加严重限制。我们的结果在某种程度上解决了这个问题，因为我们涵盖了一个非常规货币政策的时期，我们通过在我们的 BVAR 中包括短期和长期利率来明确考虑这一点，并且在进行的稳健性检验中，使用了一个可能有助于捕捉非常规政策的影子利率。然而，对监管/监督与货币政策干预之间互动的更直接的实证分析——例如，在类似于 Davidson 和 Moccero（2024）的非线性设置中——将是研究的一个重要进展。其次，未来的研究可以考察其他国家，并通过纳入制造业、服务业和建筑业的数据来确定哪些行业受到资本要求增加的影响最大。我们的框架确实可以轻松扩展，甚至可以扩展到多国/多行业模型。第三，纳入存款流动和利率将允许对银行绩效及相关宏观经济影响进行更全面的分析。最后，将样本扩展到涵盖 Covid-19 大流行（Altavilla et al. 2023）

^{47}

及随后的高通胀时期，当时银行的风险感知严重影响了贷款动态（Auer and Conti 2024），将拓宽我们对审慎政策与银行贷款之间关系的认识。

Acknowledgments致谢

We thank the participants to seminars at NY FED, University of Pennsylvania, Universitat Pompeu Fabra, Banca d’Italia and the ECB, the 2024 International Conference on Macroeconomic Analysis and International Finance, the 2022 ASSA Meeting, the 2019 Annual Conference on Real-Time Data, Methods and Applications, the 2nd Annual Workshop of the ESCB Research Cluster 3 on Financial Stability, Macroprudential Regulation and Microprudential Supervision, the 2018 Annual Congress of the EEA, the 2018 IAAE Annual Conference, the 1st Vienna Workshop on Economic Forecasting, and the 5th Research Workshop of the MPC Task Force on Banking Analysis for Monetary Policy. Without any implications, we are particularly grateful to Florin Bilbiie, Christian Brownlees, Marianna Caccavaio, Dean Croushore, Alessio De Vincenzo, Jesus Fernandez Villaverde, Massimo Franchi, Domenico Giannone, Joni Heikkinen, Michele Lenza, JeanStéphane Mésonnier, Stefano Neri, Michael O’Grady, Gert Peersman, Giorgio Primiceri, Juan Rubio-Ramirez, Frank Schorfeide, Alessandro Secchi, Stefano Siviero, Belinda Tracey, and Fabrizio Venditti for useful comments and suggestions. A first draft of this paper was written when Antonio M. Conti was a Visiting Scholar at the Department of Economics of Pennsylvania University, whose hospitality is gratefully acknowledged. All errors are of the authors. The views expressed herein are those of the authors and do not necessarily reflect those of the Banca d’Italia or the Eurosystem.
我们感谢在纽约联邦储备银行、宾夕法尼亚大学、庞培法布拉大学、意大利银行和欧洲中央银行参加研讨会的参与者，感谢 2024 年国际宏观经济分析与国际金融会议、2022 年 ASSA 会议、2019 年实时数据、方法与应用年会、欧洲中央银行系统研究集群 3 关于金融稳定、宏观审慎监管和微观审慎监督的第二届年会、2018 年欧洲经济协会年会、2018 年国际应用经济学协会年会、第一届维也纳经济预测研讨会，以及货币政策银行分析工作组的第五届研究研讨会。我们特别感谢 Florin Bilbiie、Christian Brownlees、Marianna Caccavaio、Dean Croushore、Alessio De Vincenzo、Jesus Fernandez Villaverde、Massimo Franchi、Domenico Giannone、Joni Heikkinen、Michele Lenza、JeanStéphane Mésonnier、Stefano Neri、Michael O’Grady、Gert Peersman、Giorgio Primiceri、Juan Rubio-Ramirez、Frank Schorfeide、Alessandro Secchi、Stefano Siviero、Belinda Tracey 和 Fabrizio Venditti 提供的有用意见和建议。本文的初稿是在安东尼奥·M·孔蒂担任宾夕法尼亚大学经济系访问学者期间撰写的，特此感谢其热情款待。所有错误均由作者承担。本文所表达的观点仅代表作者个人，并不一定反映意大利银行或欧元体系的观点。

Data Availability Statement
数据可用性声明

The authors state that most of the banking variables used in this study are confidential and cannot be shared (e.g., Tier 1 equity, R.W.A., net interest income and loan loss provisions). The other data are available from the corresponding author upon reasonable request.
作者声明，本研究中使用的大多数银行变量是机密的，无法共享（例如，一级资本、风险加权资产、净利息收入和贷款损失准备）。其他数据可根据合理请求向通讯作者获取。

Endnotes尾注

^{1}

The differences stem from several dimensions: the methodology, the type of data (micro vs. macro), the country or the sample considered.

^{1}

差异源于几个方面：方法论、数据类型（微观与宏观）、考虑的国家或样本。

^{2}

Accornero et al. (2017) take advantage of the launch of the Banking Union and the SSM and the European Central Bank’s (ECB) Comprehensive Assessment (CA) to assess the impact of an increase in nonperforming loans on credit supply, using a panel of Italian banks.

^{2}

Accornero et al. (2017) 利用银行联盟和单一监管机制（SSM）以及欧洲中央银行（ECB）的全面评估（CA）的启动，评估不良贷款增加对信贷供应的影响，使用了一组意大利银行的数据。

^{3}

A related issue is that theory is not unanimous on how key macroeconomic variables should react to changes in bank capital requirements: in particular, the response of inflation is far from settled, with some papers suggesting a positive co-movement between prices and economic activity, and thus assessing bank capital as a demand-side disturbance, while others instead reach the opposite conclusion (on this issue, see the evidence reviewed in Gambetti and Musso 2017).

^{3}

一个相关的问题是，理论对关键宏观经济变量应如何对银行资本要求的变化作出反应并不一致：特别是，通货膨胀的反应远未确定，一些论文建议价格与经济活动之间存在正相关关系，因此将银行资本评估为需求侧干扰，而另一些论文则得出相反的结论（关于这个问题，参见 Gambetti 和 Musso 2017 年回顾的证据）。

^{4}

For example, in Noss and Toffano (2016), no significant effect on economic activity is found, and in Kanngiesser et al. (2020), the effect on real GDP is barely significant, lasting only one quarter.

^{4}

例如，在 Noss 和 Toffano（2016）中，没有发现对经济活动的显著影响，而在 Kanngiesser 等人（2020）中，对实际 GDP 的影响几乎不显著，仅持续一个季度。

^{5}

Altavilla, Canova, and Ciccarelli (2020) perform a similar exercise to obtain the counterfactual pattern of the EONIA rate, government bond yields, and bank bond yields in the absence of the ECB’s unconventional monetary policy interventions.

^{5}

Altavilla, Canova, 和 Ciccarelli (2020) 进行了一项类似的研究，以获得在没有欧洲中央银行非常规货币政策干预的情况下，EONIA 利率、政府债券收益率和银行债券收益率的反事实模式。

^{6}

An alternative to these vintages could be given by private sector forecasts, such as the Consensus. Unfortunately, only real GDP and consumer prices are available.

^{6}

这些年份的替代方案可以由私营部门的预测提供，例如共识。不幸的是，只有实际 GDP 和消费价格可用。

^{7}

This is equivalent to what Altavilla, Canova, and Ciccarelli (2020) and Giannone et al. (2012), in their respective settings, argue in relation to estimating the size of the ECB’s unconventional monetary policy interventions based on negative deposit facility rates and the evolution of the Eurosystem’s balance sheet.

^{7}

这相当于 Altavilla、Canova 和 Ciccarelli (2020) 以及 Giannone 等人 (2012) 在各自的背景下，关于基于负存款便利利率和欧元体系资产负债表演变来估计欧洲央行非常规货币政策干预规模的论点。

^{8}

In this regard, this paper is related to the literature which uses narrative analysis to study monetary policy and financial shocks (e.g., Romer and Romer 2004, 2017; Eickmeier, Kolb, and Prieto 2018). One difference with those papers is that we do not rely on local projections (Jorda 2005).

^{8}

在这方面，本文与使用叙事分析研究货币政策和金融冲击的文献相关（例如，Romer 和 Romer 2004, 2017；Eickmeier, Kolb 和 Prieto 2018）。与这些论文的一个不同之处在于，我们不依赖于局部预测（Jorda 2005）。

^{9}

The Basel III reform was definitively approved only in September 2010 and scheduled to be introduced from 2013. However, the details of the reform were largely anticipated by banks, which started to increase their capital buffers as soon as the discussion began. Two key meetings in this respect were the G20 summits in April and November 2009, where the leaders committed to completing a global reform of prudential regulation (Banca d’Italia 2010); subsequently, in December 2009, the Basel Committee on Banking Supervision published a consultation paper with concrete proposals for capital and liquidity regulation reforms (Basel Committee on Banking Supervision 2009).

^{9}

巴塞尔协议 III 的改革在 2010 年 9 月最终获得批准，并计划于 2013 年实施。然而，银行在讨论开始时就大致预见到了改革的细节，并开始增加其资本缓冲。与此相关的两个关键会议是 2009 年 4 月和 11 月的 G20 峰会，领导人承诺完成全球审慎监管改革（意大利银行 2010）；随后，在 2009 年 12 月，巴塞尔银行监管委员会发布了一份咨询文件，提出了资本和流动性监管改革的具体建议（巴塞尔银行监管委员会 2009）。

^{10}

The stress test was conducted between January and July 2011. The EBA allowed specific capital actions in the first four months of 2011 to be considered in the final result; the five Italian banks participating to the test raised about 11 billion. The Capital Exercise was announced in October 2011, and banks were prescribed to cover possible shortfalls by the end of June 2012; the total shortfall identified in the exercise for the Italian banks amounted to 15 billion.

^{10}

压力测试在 2011 年 1 月至 7 月之间进行。欧洲银行管理局允许在 2011 年前四个月进行特定的资本行动以纳入最终结果；参与测试的五家意大利银行筹集了约 110 亿欧元。资本测试于 2011 年 10 月宣布，银行被要求在 2012 年 6 月底之前弥补可能的缺口；在此次测试中，意大利银行识别出的总缺口为 150 亿欧元。

^{11}

The CA was announced at the end of 2013 and completed in October 2014. For Italian banks, the results published by the ECB envisaged an aggregate capital requirement of about 3 billion; in addition, between

^{11}

CA 于 2013 年底宣布，并于 2014 年 10 月完成。对于意大利银行，欧洲央行发布的结果预计总资本需求约为 30 亿；此外，在

January and September 2014, a number of capital strengthening measures were taken, both by banks that ultimately did not pass the CA and banks that did, amounting to about 11 billion. Further capital increases were recorded in the first half of 2015.
2014 年 1 月和 9 月，采取了一系列资本增强措施，包括未能通过 CA 的银行和通过 CA 的银行，总额约为 110 亿。2015 年上半年进一步记录了资本增加。

^{12}

This methodology complements traditional sign restrictions with restrictions on the sign and contribution of the identified shock to a particular variable of interest-derived from narrative informationand provides a perspective close to that used in our scenario analysis. In particular, the flexibility of this method is such that we can impose narrative restrictions exclusively on the policy variable (in our case, the Tier 1 ratio), thus being agnostic about the variables for which we want to quantify the impact of policy actions.

^{12}

这种方法论补充了传统的符号限制，并对特定感兴趣变量的识别冲击的符号和贡献施加限制——这些限制源于叙述信息，并提供了与我们情景分析中使用的视角相近的观点。特别是，这种方法的灵活性使我们能够仅对政策变量（在我们的案例中是一级资本充足率）施加叙述限制，从而对我们希望量化政策行动影响的变量保持中立。

^{13}

In these papers, the authors follow a “two-step” approach. In the first stage, they recover estimates of “economic level of capital” by running panel regressions of bank capital on a number of macroeconomic and bank-level variables, such as measures of profitability and riskiness. Then, they recover bank-level capital shortfall/ overhang as the difference between actual and “economic level” of capital, which are summed up to obtain an aggregate measure of bank capital shocks. In the second stage, they put the bank capital shock in VAR model and use a Choleski decomposition to recover impulse responses.
在这些论文中，作者采用了“二步”方法。在第一阶段，他们通过对银行资本与多种宏观经济和银行层面变量（如盈利能力和风险性指标）进行面板回归，恢复“经济资本水平”的估计。然后，他们将实际资本与“经济水平”资本之间的差异恢复为银行层面的资本短缺/过剩，并将其汇总以获得银行资本冲击的总体度量。在第二阶段，他们将银行资本冲击放入 VAR 模型中，并使用 Choleski 分解来恢复脉冲响应。

^{14}

A recent contribution uses a narrative approach to build an index of aggregate tightenings in regulatory US bank capital requirements from 1979 to 2008 (Eickmeier, Kolb, and Prieto 2018).

^{14}

最近的一项研究采用叙述方法构建了一个从 1979 年到 2008 年美国银行资本要求的整体收紧指数（Eickmeier, Kolb, and Prieto 2018）。

^{15}

A different approach, combining structural identification and conditional forecasts, has been recently proposed by Antolín-Díaz, Petrella, and Rubio-Ramírez (2021). This methodology helps researchers interested in understanding what are the shocks driving the conditioning set variables, being therefore complementary to the techniques adopted in our paper, where we instead aim to construct a counterfactual policy variable. See Sections 4.4 and 5 for more details.

^{15}

Antolín-Díaz、Petrella 和 Rubio-Ramírez（2021）最近提出了一种不同的方法，结合了结构识别和条件预测。这种方法帮助研究人员理解驱动条件集变量的冲击，因此与我们论文中采用的技术是互补的，我们的目标是构建一个反事实政策变量。有关更多细节，请参见第 4.4 节和第 5 节。

^{16}

The SSM was established by the EU Council Regulation 1024/2013 and consists of the joint exercise of supervisory tasks and powers vis-à-vis the banks on behalf of the ECB (with the newly created Supervisory Board) and of euro-area supervisory authorities. The ECB supervises the “Significant Institutions” (SIs) directly, while the other banks (“Less significant Institutions”, LSIs) are supervised by the national authorities (though the SSM retains the ultimate power also on these institutions). The classification as SIs/LSIs is based on total assets, size of total assets relative to country GDP, and scope of cross-border activities.

^{16}

SSM 是根据欧盟理事会条例 1024/2013 建立的，负责代表欧洲中央银行（ECB）和欧元区监管机构共同行使对银行的监管任务和权力（由新成立的监管委员会负责）。欧洲中央银行直接监管“重要机构”（SIs），而其他银行（“不太重要机构”，LSIs）则由国家当局监管（尽管 SSM 对这些机构也保留最终权力）。SIs/LSIs 的分类基于总资产、总资产相对于国家 GDP 的规模以及跨境活动的范围。

^{17}

The Tier 1 capital ratio is a key regulatory measure of a bank’s capital adequacy. The numerator (Tier 1 capital) consists of Common Equity Tier 1 (CET1)—that is, common shares, stock surpluses resulting from the issue of common shares, retained earnings, common shares issued by subsidiaries and held by third parties, accumulated other comprehensive income (AOCI)—and Additional Tier 1 capital (AT1)—which includes instruments that are not common equity but are eligible to be included in this tier, such as contingent convertible or hybrid securities, which have a perpetual term and can be converted into equity when a trigger event occurs. RWAs are total bank assets (including off-balance-sheet exposures) weighted according to their riskiness.

^{17}

一级资本充足率是衡量银行资本充足性的关键监管指标。分子（一级资本）由普通股一级资本（CET1）组成——即普通股、因发行普通股而产生的股票溢价、留存收益、由子公司发行并由第三方持有的普通股、累计其他综合收益（AOCI）——以及附加一级资本（AT1）——包括不属于普通股但符合该级别的工具，例如有条件可转换或混合证券，这些证券具有永久期限，并且在触发事件发生时可以转换为股权。风险加权资产（RWA）是根据其风险性加权的银行总资产（包括表外风险敞口）。

^{18}

In the two other periods mentioned, the reduction in the ratio in the second half of the 1990s was the result of a rapid expansion in RWAs (with annual growth rates of around

10 %

) and a modest growth in Tier 1 capital. In the 2000s, up to the onset of the financial crisis, both the numerator and the denominator grew substantially, though at a similar pace.
在提到的另外两个时期中，1990 年代后半期比例的下降是由于风险加权资产（RWAs）快速扩张（年增长率约为

10 %

）和一级资本的适度增长所致。在 2000 年代，直到金融危机的爆发，分子和分母都大幅增长，尽管增速相似。

^{19}

See https://www.bis.org/press/p081120.htm.

^{19}

请参见 https://www.bis.org/press/p081120.htm。

^{20}

Key provisions of the reform included the following: (i) a minimum of

4.5 %

for the banks’ CET 1, up from the level of

2 %

dictated in the Basel II framework; (ii) a minimum overall Tier 1 ratio requirement of

6 %

, up from

4 %

in Basel II; (iii) the introduction of a minimum requirement for Tier 1 capital as a ratio to non-RWAs (the leverage ratio, set at

3 %

); (iv) the introduction of the Capital Conservation Buffer and the Countercyclical Capital Buffer, additional capital requirements with macroprudential purposes; (v) the introduction of minimum liquidity standards (the LCR and the NSFR); and (vi) the introduction of supplemental Pillar 2 requirements addressing, among other things, firm governance and risk management and interest rate risk.

^{20}

改革的关键条款包括以下内容：(i) 银行的 CET 1 最低要求为

4.5 %

，高于巴塞尔 II 框架规定的

2 %

水平；(ii) 整体一级资本比率最低要求为

6 %

，高于巴塞尔 II 中的

4 %

；(iii) 引入对非风险加权资产的一级资本最低要求（杠杆比率，设定为

3 %

）；(iv) 引入资本保护缓冲和逆周期资本缓冲，作为具有宏观审慎目的的额外资本要求；(v) 引入最低流动性标准（LCR 和 NSFR）；以及(vi) 引入补充第二支柱要求，涉及公司治理、风险管理和利率风险等方面。

^{21}

The test included also one bank from Norway.

^{21}

测试还包括一家来自挪威的银行。

^{22}

All banks passed the test.

^{22}

所有银行都通过了测试。

^{23}

Three banks raised additional own funds (Unicredit, Banco Popolare, and UBI Banca); the capital strengthening of Banca Monte dei Paschi di Siena required an intervention by the government, by about 2 billion euro (Banca d’Italia 2012).

^{23}

三家银行增加了额外的自有资金（Unicredit、Banco Popolare 和 UBI Banca）；蒙特帕斯基银行的资本增强需要政府约 20 亿欧元的干预（意大利银行 2012）。

^{24}

On the CA and its outcome, see https://www.bankingsupervision. europa.eu/press/pr/date/2014/html/sr141026.en.html and https:// www.bankingsupervision.europa.eu/ecb/pub/pdf/aggregatereport onthecomprehensiveassessment201410.en.pdf?68911b281b9d831 540bb474c334437e7.

^{24}

关于 CA 及其结果，请参见 https://www.bankingsupervision.europa.eu/press/pr/date/2014/html/sr141026.en.html 和 https://www.bankingsupervision.europa.eu/ecb/pub/pdf/aggregatereportonthecomprehensiveassessment201410.en.pdf?68911b281b9d831540bb474c334437e7。

^{25}

See Banca d’Italia (2014) for a detailed description of the results for Italian banks. The capital increases were undertaken both by banks that passed the CA and by banks for which a shortfall was identified.

^{25}

请参阅意大利银行（Banca d’Italia）（2014）以获取意大利银行结果的详细描述。资本增加是由通过 CA 的银行和被识别出短缺的银行共同进行的。

^{26}

The description of the empirical framework largely draws on Aastveit et al. (2017) and Clark and McCracken (2014).

^{26}

实证框架的描述主要基于 Aastveit et al. (2017) 和 Clark and McCracken (2014)。

^{27}

Again, further details can be found in the Supporting Information Appendix A.

^{27}

再次，更多细节可以在支持信息附录 A 中找到。

^{28}

For robustness, we have ran all the analysis described in the paper by replacing the 3 -month Euribor rate with the measure of the so-called shadow rate developed by Krippner (2013). The shadow rate provides a measure of the monetary stance at the Zero Lower Bound by parsimoniously summarizing the information about the monetary stance that is embedded in the term structure of interest rates. Therefore, movements of the shadow rate tend to be broadly correlated with the events related to the adoption of nonstandard monetary policy measures (Krippner, 2013). In addition to Krippner (2013), see Wu and Xia (2016) for a description of the shadow rates estimation; von Borstel, Eickmeier, and Krippner (2016), Conti (2017), and Albertazzi, Nobili, and Signoretti (2021) provide empirical applications in monetary analysis. All of our results are not affected by the use of this alternative measure and are available upon request.
为了稳健性，我们已通过用 Krippner（2013）开发的所谓影子利率替代 3 个月 Euribor 利率，进行了论文中描述的所有分析。影子利率通过简洁地总结嵌入利率期限结构中的货币立场信息，提供了在零下限的货币立场的度量。因此，影子利率的变动往往与与采用非常规货币政策措施相关的事件广泛相关（Krippner，2013）。除了 Krippner（2013），请参见 Wu 和 Xia（2016）关于影子利率估计的描述；von Borstel、Eickmeier 和 Krippner（2016）、Conti（2017）以及 Albertazzi、Nobili 和 Signoretti（2021）提供了货币分析中的实证应用。我们所有的结果不受使用这种替代度量的影响，并可根据请求提供。

^{29}

Bank lending volume is the outstanding amount of loans extended by the Italian Monetary and Financial Institutions (MFIs) to NFCs and HHs resident in Italy, adjusted for including the impact of securitizations and reclassifications; lending rate is the average rate on the stock of loans with maturity up to 1 year, for NFCs, and on the flow of new loans in a given quarter, for HHs.

^{29}

银行贷款量是意大利货币和金融机构（MFI）向居住在意大利的非金融公司（NFC）和家庭（HH）发放的未偿还贷款总额，已调整以包括证券化和重新分类的影响；贷款利率是针对非金融公司（NFC）到期不超过 1 年的贷款存量的平均利率，以及针对家庭（HH）在特定季度的新贷款流量的平均利率。

^{30}

Default rates are the seasonally adjusted quarterly flow of new bad loans of Italian banks, expressed as a ratio to the stock of outstanding loans at the beginning of the period.

^{30}

默认利率是意大利银行新坏账的季节性调整季度流量，以期初未偿贷款余额的比率表示。

^{31}

Each of the bank income statement variables is the four-quarter moving sum of the item reported by Italian banks in aggregate.

^{31}

每个银行收入报表变量是意大利银行汇总报告的项目的四个季度移动总和。

^{32}

The index is the sectoral Index FTSE Italia All-Share Banks, from Borsa Italiana.

^{32}

指数是来自意大利证券交易所的行业指数 FTSE 意大利全股银行指数。

^{33}

The Tier 1 ratio is the total amount of Tier 1 equity of Italian banks divided by RWAs.

^{33}

一级资本充足率是意大利银行一级资本总额与风险加权资产（RWA）的比率。

^{34}

In particular, for each variable, we estimate an

AR (1)

model via the Box-Jenkins techniques over the entire sample period and set the prior at the value of the estimated coefficient.

^{34}

特别是，对于每个变量，我们通过 Box-Jenkins 技术在整个样本期间估计一个

AR (1)

模型，并将先验设置为估计系数的值。

^{35}

This procedure follows Favara and Giordani (2009) who evaluate the incremental predictive content of money for the three main variables of the New Keynesian model (output, prices, and interest rates). It can be considered as a multivariate Granger causality test, which allows for testing the relevance of a (number of) variable(s) for forecasting the remaining set of endogenous time series included in the VAR model.

^{35}

本程序遵循 Favara 和 Giordani（2009）的研究，他们评估了货币对新凯恩斯模型三个主要变量（产出、价格和利率）的增量预测内容。它可以被视为一种多变量格兰杰因果关系检验，允许测试一个（或多个）变量对预测 VAR 模型中包含的其余内生时间序列的相关性。

^{36}

It is also possible to build forecasts of some

k

variables of interest

Y_{1}

, …,

Y_{k}

on a particular path of some structural shocks of interest over the forecast horizon. This framework can be labeled “conditional-onshocks forecast” (see, e.g., Baumeister and Kilian 2014; Antolín-Díaz, Petrella, and Rubio-Ramírez 2021). There is a key difference between this methodology and the “conditional-on-variables” one: While the former requires the estimated parameters of the structural form of the VAR model and thus identifying assumptions for the shocks of interest, the latter relies on the reduced-form parameters of the VAR only.

^{36}

还可以在预测区间内基于某些结构性冲击的特定路径构建一些

k

变量的预测

Y_{1}

, …,

Y_{k}

。这个框架可以被称为“条件性冲击预测”（参见，例如，Baumeister 和 Kilian 2014；Antolín-Díaz, Petrella 和 Rubio-Ramírez 2021）。该方法与“条件性变量”方法之间有一个关键区别：前者需要 VAR 模型结构形式的估计参数，因此需要识别感兴趣冲击的假设，而后者仅依赖于 VAR 的简化形式参数。

^{37}

Waggoner and Zha (1999) developed a Gibbs sampling algorithm that provides the exact finite-sample distribution of the conditional forecasts, by taking the conditions into account when sampling the VAR coefficients. We abstract from this method because it is computationally very intensive in medium-scale models. Moreover, Clark and McCracken (2014) and Aastveit et al. (2017) found that, in smaller models, the various methods provide extremely similar results.

^{37}

Waggoner 和 Zha (1999) 开发了一种 Gibbs 采样算法，该算法通过在采样 VAR 系数时考虑条件，提供了条件预测的精确有限样本分布。我们抽象出这种方法，因为它在中等规模模型中计算量非常大。此外，Clark 和 McCracken (2014) 以及 Aastveit 等 (2017) 发现，在较小的模型中，各种方法提供的结果极为相似。

^{38}

For robustness, we also run conditional forecasts including other items of bank profitability in the conditioning set (loan loss provisions, noninterest income, operational expenses), obtaining very similar results (not reported and available upon request).

^{38}

为了稳健性，我们还进行条件预测，包括银行盈利能力的其他项目在条件集中（贷款损失准备金、非利息收入、运营费用），获得非常相似的结果（未报告，可根据请求提供）。

^{39}

To the extent that the rise in Tier 1 capital has an impact on the some of its own macro-financial and banking drivers, then using the realized evolution of the latter to derive a “no-policy scenario” would give rise to distorted estimations. We thank the Associate Editor for pointing out to this issue.

^{39}

在一级资本上升对其自身的一些宏观金融和银行驱动因素产生影响的情况下，利用后者的实际演变来推导“无政策情景”将导致扭曲的估计。我们感谢副编辑指出这个问题。

^{40}

More specifically, the path imposed for these variables is the one consistent with the counterfactual scenario for the Tier 1 ratio, namely: (i) their unconditional forecast in the VAR, when we use the unconditional forecast from the VAR as the counterfactual Tier 1 ratio; (ii) their expected values in the corresponding vintage of the Eurosystem projections, when we use these projections to obtain the counterfactual Tier 1 ratio.

^{40}

更具体地说，这些变量所施加的路径与一级资本比率的反事实情景是一致的，即：(i) 当我们使用 VAR 的无条件预测作为反事实一级资本比率时，它们在 VAR 中的无条件预测；(ii) 当我们使用这些预测来获得反事实一级资本比率时，它们在相应年份的欧洲系统预测中的期望值。

^{41}

The latter finding is consistent with the possibility that the effects of credit supply tightening may be be associated with both aggregate demand and aggregate supply effects. The latter case may, in turn, be related to mechanisms such as those in Gilchrist et al. (2017), where firms which are credit constrained and face low demand because of low growth raise prices instead of lowering them, in order to preserve internal liquidity.

^{41}

后者的发现与信贷供应收紧的影响可能与总需求和总供应效应相关的可能性一致。后者的情况可能又与 Gilchrist 等人（2017）中的机制有关，在这些机制中，受到信贷限制且由于低增长面临低需求的企业提高价格而不是降低价格，以保持内部流动性。

^{42}

Albertazzi et al. (2014), using reduced-form single-equation models, document a change in the estimated relationships between the sovereign spread and a number of banking variables in Italy, including lending conditions and profitability, since 2011Q3.

^{42}

Albertazzi 等人 (2014) 使用简化形式的单方程模型，记录了自 2011 年第三季度以来，意大利主权利差与多项银行变量之间估计关系的变化，包括贷款条件和盈利能力。

^{43}

The in-sample exercise should be considered more as of a check on the ability of the model in capturing the dynamics of Tier 1 ratio when exploiting all the available information (especially the increase in its volatility in 2014-2015) than a proper method for computing IRFs (see also Aastveit et al. 2017). Indeed, using coefficients estimated over the full sample would introduce a potential inconsistency in our approach when estimating the (loose) IRFs, as this procedure would implicitly assume that information from future developments of variables under evaluation is used to run current forecasts.

^{43}

样本内的练习应被视为对模型在利用所有可用信息（特别是 2014-2015 年波动性增加）时捕捉一级资本比率动态能力的检查，而不是计算脉冲响应函数（IRFs）的适当方法（另见 Aastveit 等，2017）。实际上，使用在整个样本上估计的系数会在我们估计（宽松的）脉冲响应函数时引入潜在的不一致性，因为该程序会隐含地假设使用了被评估变量未来发展的信息来进行当前预测。

^{44}

See the BLS website at https://www.ecb.europa.eu/stats/ecb_surve ys/bank_lending_survey/html/index.en.html.

^{44}

请访问 BLS 网站 https://www.ecb.europa.eu/stats/ecb_surve ys/bank_lending_survey/html/index.en.html。

^{45}

This is the so called “missing inflation” period; see, for example, Bobeica and Jarociński (2019).

^{45}

这就是所谓的“缺失通胀”时期；例如，参见 Bobeica 和 Jarociński (2019)。

^{46}

We download the series of the US Tier 1 capital ratio from the NY FED website (https://www.newyorkfed.org/research/banking_resea rch/quarterly_trends.html).

^{46}

我们从纽约联邦储备银行网站 (https://www.newyorkfed.org/research/banking_resea rch/quarterly_trends.html) 下载美国一级资本充足率的系列数据。

^{47}

In this regard, preliminary findings obtained extending the dataset up to 2023:Q4 show a very good forecasting performance of our baseline model for the evolution of Tier 1 ratio over the period 2020:Q1-2023:Q4.

^{47}

在这方面，扩展数据集至 2023 年第四季度的初步发现显示，我们的基线模型在 2020 年第一季度至 2023 年第四季度期间对一级资本充足率的演变具有非常好的预测表现。

References参考文献

Aastveit, K. A., A. Carriero, T. E. Clark, and M. Marcellino. 2017. “Have Standard VARs Remained Stable Since the Crisis?.” Journal of Applied Econometrics 32, no. 5: 931-951.
Aastveit, K. A., A. Carriero, T. E. Clark, 和 M. Marcellino. 2017. “标准 VAR 自危机以来是否保持稳定？.” 应用计量经济学杂志 32, no. 5: 931-951.

Accornero, M., P. Alessandri, L. Carpinelli, and A. M. Sorrentino. 2017. “Non-Performing Loans and the Supply of Bank Credit: Evidence From Italy.” 374, Bank of Italy, Economic Research and International Relations Area.
Accornero, M., P. Alessandri, L. Carpinelli, 和 A. M. Sorrentino. 2017. “不良贷款与银行信贷供给：来自意大利的证据。” 374, 意大利银行，经济研究与国际关系领域。

Aiyar, S., C. W. Calomiris, and T. Wieladek. 2016. “How Does Credit Supply Respond to Monetary Policy and Bank Minimum Capital Requirements?” European Economic Review 82: 142-165.
Aiyar, S., C. W. Calomiris, 和 T. Wieladek. 2016. “信贷供应如何响应货币政策和银行最低资本要求？” 欧洲经济评论 82: 142-165.

Albertazzi, U., A. Nobili, and F. M. Signoretti. 2021. “The Bank Lending Channel of Conventional and Unconventional Monetary Policy.” Journal of Money, Credit and Banking 53, no. 2-3: 261-299.
Albertazzi, U., A. Nobili, 和 F. M. Signoretti. 2021. “传统与非常规货币政策的银行贷款渠道。”《货币、信贷与银行学报》53, no. 2-3: 261-299.

Albertazzi, U., T. Ropele, G. Sene, and F. M. Signoretti. 2014. “The Impact of the Sovereign Debt Crisis on the Activity of Italian Banks.” Journal of Banking & Finance 46, no. C: 387-402.
Albertazzi, U., T. Ropele, G. Sene, 和 F. M. Signoretti. 2014. “主权债务危机对意大利银行活动的影响。”《银行与金融杂志》46 卷, C 期: 387-402.

Altavilla, C., F. Barbiero, M. Boucinha, and L. Burlon. 2023. “The Great Lockdown: Pandemic Response Policies and Bank Lending Conditions.” European Economic Review 156: 104478. https://ideas.repec.org/a/eee/ eecrev/v156y2023ics0014292123001071.html.
Altavilla, C., F. Barbiero, M. Boucinha, 和 L. Burlon. 2023. “大封锁：疫情应对政策与银行贷款条件。” 欧洲经济评论 156: 104478. https://ideas.repec.org/a/eee/eecrev/v156y2023ics0014292123001071.html.

Altavilla, C., F. Canova, and M. Ciccarelli. 2020. “Mending the Broken Link: Heterogeneous Bank Lending Rates and Monetary Policy PassThrough.” Journal of Monetary Economics 110: 81-98. http://www.scien cedirect.com/science/article/pii/S0304393219300030.
Altavilla, C., F. Canova, 和 M. Ciccarelli. 2020. “修复断裂的联系：异质银行贷款利率与货币政策传导。” 《货币经济学杂志》 110: 81-98. http://www.sciencedirect.com/science/article/pii/S0304393219300030.

Altavilla, C., D. Giannone, and M. Lenza. 2016. “The Financial and Macroeconomic Effects of the OMT Announcements.” International Journal of Central Banking 12, no. 3: 29-57.
Altavilla, C., D. Giannone, 和 M. Lenza. 2016. “OMT 公告的金融和宏观经济影响。” 国际中央银行学杂志 12, no. 3: 29-57.

Angelini, P., L. Clerc, V. Curdia, et al. 2011. “Basel III: Long-Term Impact on Economic Performance and Fluctuations.” 87, Bank of Italy, Economic Research and International Relations Area.
Angelini, P., L. Clerc, V. Curdia, et al. 2011. “巴塞尔协议 III：对经济表现和波动的长期影响。” 87, 意大利银行，经济研究与国际关系领域。

Antolín-Díaz, J., I. Petrella, and J. F. Rubio-Ramírez. 2021. “Structural Scenario Analysis With SVARs.” Journal of Monetary Economics 117: 798-815. http://www.sciencedirect.com/science/article/pii/S0304 393220300751.
Antolín-Díaz, J., I. Petrella, 和 J. F. Rubio-Ramírez. 2021. “使用 SVAR 的结构情景分析。”《货币经济学杂志》117: 798-815. http://www.sciencedirect.com/science/article/pii/S0304 393220300751.

Antolín-Díaz, J., and J. F. Rubio-Ramírez. 2018. “Narrative Sign Restrictions for SVARs.” American Economic Review 108, no. 10: 2802-2829. https://ideas.repec.org/a/aea/aecrev/v108y2018i10p280229.html.
Antolín-Díaz, J., 和 J. F. Rubio-Ramírez. 2018. “SVAR 的叙事信号限制.” 美国经济评论 108, no. 10: 2802-2829. https://ideas.repec.org/a/aea/aecrev/v108y2018i10p280229.html.

Auer, S., and A. M. Conti. 2024. “Bank Lending in an Unprecedented Monetary Tightening Cycle: Evidence From the Euro Area.” 856, Bank of Italy, Economic Research and International Relations Area. https:// ideas.repec.org/p/bdi/opques/qef_856_24.html.
Auer, S., 和 A. M. Conti. 2024. “在前所未有的货币紧缩周期中的银行贷款：来自欧元区的证据。” 856, 意大利银行, 经济研究与国际关系领域. https:// ideas.repec.org/p/bdi/opques/qef_856_24.html.

Banbura, M., D. Giannone, and M. Lenza. 2015. “Conditional Forecasts and Scenario Analysis With Vector Autoregressions for Large CrossSections.” International Journal of Forecasting 31, no. 3: 739-756.
Banbura, M., D. Giannone, 和 M. Lenza. 2015. “使用大横截面的向量自回归进行条件预测和情景分析。” 国际预测杂志 31, no. 3: 739-756.

Banbura, M., D. Giannone, and L. Reichlin. 2010. “Large Bayesian Vector Auto Regressions.” Journal of Applied Econometrics 25, no. 1: 71-92.
Banbura, M., D. Giannone, 和 L. Reichlin. 2010. “大型贝叶斯向量自回归。” 应用计量经济学杂志 25, no. 1: 71-92.

Banca d’Italia. 2010. “Annual Report for 2009.” Technical Report May, Banca d’Italia, Roma.
意大利银行。2010。“2009 年年度报告。”技术报告，5 月，意大利银行，罗马。

Banca d’Italia. 2012. “Financial Stability Report N.4.” Technical Report November. Banca d’Italia, Roma.
意大利银行。2012。“金融稳定报告 N.4。” 技术报告 11 月。意大利银行，罗马。
Banca d’Italia. 2014. “Results of the Comprehensive Assessment.” Press Releases October. Banca d’Italia, Roma. https://www.bancaditalia.it/ media/comunicati/documenti/2014-02/cs_261014_en.pdf?language_ id%3D1.
意大利银行。2014。“综合评估结果。”新闻稿十月。意大利银行，罗马。https://www.bancaditalia.it/ media/comunicati/documenti/2014-02/cs_261014_en.pdf?language_ id%3D1.

Basel Committee on Banking Supervision. 2009. “Strengthening the Resilience of the Banking Sector.” Consultative Document December. BIS, Basel.
巴塞尔银行监管委员会。2009。“加强银行业的韧性。”咨询文件，12 月。国际清算银行，巴塞尔。

Basel Committee on Banking Supervision. 2010. “An Assessment of the Long-Term Impact of Stronger Capital and Liquidity Requirements.” Consultative Document August. BIS, Basel.
巴塞尔银行监管委员会。2010。“对更强资本和流动性要求的长期影响评估。”咨询文件八月。国际清算银行，巴塞尔。

Basel Committee on Banking Supervision. 2015. “Eighth Progress Report on Adoption of the Basel Regulatory Framework.” Technical Report December. BIS, Basel.
巴塞尔银行监管委员会。2015。“巴塞尔监管框架采纳的第八次进展报告。”技术报告，12 月。国际清算银行，巴塞尔。

Bassett, W. F., S. J. Lee, and T. P. Spiller. 2015. “Estimating Changes in Supervisory Standards and Their Economic Effects.” Journal of Banking & Finance 60: 21-43.
Bassett, W. F., S. J. Lee, 和 T. P. Spiller. 2015. “估计监管标准的变化及其经济影响。” 《银行与金融杂志》 60: 21-43.

Baumeister, C., and L. Kilian. 2014. “Real-Time Analysis of Oil Price Risks Using Forecast Scenarios.” IMF Economic Review 62, no. 1: 119-145.
Baumeister, C., 和 L. Kilian. 2014. “使用预测情景对油价风险进行实时分析。” 国际货币基金组织经济评论 62, no. 1: 119-145.

Bernanke, B. S., and M. Gertler. 1995. “Inside the Black Box: The Credit Channel of Monetary Policy Transmission.” Journal of Economic Perspectives 9, no. 4: 27-48. https://ideas.repec.org/a/aea/jecper/v9y19 95i4p27-48.html.
伯南克, B. S., 和 M. Gertler. 1995. “黑箱内部：货币政策传导的信贷渠道。” 经济学视角杂志 9, no. 4: 27-48. https://ideas.repec.org/a/aea/jecper/v9y1995i4p27-48.html.

Berrospide, J. M., and R. M. Edge. 2010. “The Effects of Bank Capital on Lending: What Do We Know, and What Does It Mean?” International Journal of Central Banking 6, no. 4: 5-54.
Berrospide, J. M., 和 R. M. Edge. 2010. “银行资本对贷款的影响：我们知道什么，这意味着什么？”《国际中央银行学杂志》6, 第 4 期: 5-54.
Bobeica, E., and M. Jarociński. 2019. “Missing Disinflation and Missing Inflation: A VAR Perspective.” International Journal of Central Banking 15, no. 1: 199-232. https://ideas.repec.org/a/ijc/ijcjou/y2019q1a5.html.
Bobeica, E., 和 M. Jarociński. 2019. “缺失的去通胀和缺失的通胀：一个 VAR 视角。” 国际中央银行学杂志 15, no. 1: 199-232. https://ideas.repec.org/a/ijc/ijcjou/y2019q1a5.html.

Bofondi, M., and T. Ropele. 2011. “Macroeconomic Determinants of Bad Loans: Evidence From Italian Banks.” 89, Bank of Italy, Economic Research and International Relations Area.
Bofondi, M., 和 T. Ropele. 2011. “不良贷款的宏观经济决定因素：来自意大利银行的证据。” 89, 意大利银行, 经济研究与国际关系领域。

Boivin, J., M. P. Giannoni, and D. Stevanovic. 2020. “Dynamic Effects of Credit Shocks in a Data-Rich Environment.” Journal of Business & Economic Statistics 38, no. 2: 272-284.
Boivin, J., M. P. Giannoni, 和 D. Stevanovic. 2020. “在数据丰富环境中信用冲击的动态效应。”《商业与经济统计杂志》38, no. 2: 272-284.

Caruso, A., L. Reichlin, and G. Ricco. 2019. “Financial and Fiscal Interaction in the Euro Area Crisis: This Time Was Different.” European Economic Review 119, no. C: 333-355.
Caruso, A., L. Reichlin, 和 G. Ricco. 2019. “欧元区危机中的金融与财政互动：这次不同。” 欧洲经济评论 119, no. C: 333-355.

Clark, T. E., and M. W. McCracken. 2014. “Evaluating Conditional Forecasts From Vector Autoregressions.” 1413, Federal Reserve Bank of Cleveland.
克拉克，T. E.，和 M. W. 麦克拉肯。2014。“评估向量自回归的条件预测。” 1413，克利夫兰联邦储备银行。

Cogley, T., and T. J. Sargent. 2005. “Drift and Volatilities: Monetary Policies and Outcomes in the Post WWII U.S.” Review of Economic Dynamics 8, no. 2: 262-302.
Cogley, T., 和 T. J. Sargent. 2005. “漂移与波动性：二战后美国的货币政策与结果。” 经济动态评论 8, no. 2: 262-302.

Conti, A. M. 2017. “Has the FED Fallen Behind the Curve? Evidence From VAR Models.” Economics Letters 159, no. C: 164-168.
Conti, A. M. 2017. “美联储是否落后于曲线？来自 VAR 模型的证据。”经济学快报 159, no. C: 164-168.

Conti, A. M., A. Nobili, and F. M. Signoretti. 2023. “Bank Capital Requirement Shocks: A Narrative Perspective.” European Economic Review 151: 104254. https://ideas.repec.org/a/eee/eecrev/v151y2023i cs0014292122001507.html.
Conti, A. M., A. Nobili, 和 F. M. Signoretti. 2023. “银行资本要求冲击：叙事视角。” 欧洲经济评论 151: 104254. https://ideas.repec.org/a/eee/eecrev/v151y2023i cs0014292122001507.html.

Dahlhaus, T., K. Hess, and A. Reza. 2018. “International Transmission Channels of U.S. Quantitative Easing: Evidence From Canada.” Journal of Money, Credit and Banking 50, no. 2-3: 545-563. https://ideas.repec. org/a/wly/jmoncb/v50y2018i2-3p545-563.html.
Dahlhaus, T., K. Hess, 和 A. Reza. 2018. “美国量化宽松的国际传导渠道：来自加拿大的证据。” 《货币、信贷与银行学杂志》 50, no. 2-3: 545-563. https://ideas.repec. org/a/wly/jmoncb/v50y2018i2-3p545-563.html.

D’Amico, S. 2016. “Discussion of “The Financial and Macroeconomic Effects of the OMT Announcements”.” International Journal of Central Banking 12, no. 3: 59-68.
D’Amico, S. 2016. “讨论《OMT 公告的金融和宏观经济影响》。”《国际中央银行杂志》12 卷，第 3 期：59-68。

Davidson, S. N., and D. N. Moccero. 2024. “The Nonlinear Effects of Banks’ Vulnerability to Capital Depletion in Euro Area Countries.” 2912, European Central Bank. https://ideas.repec.org/p/ecb/ecbwps/ 20242912.html.
Davidson, S. N., 和 D. N. Moccero. 2024. “欧元区国家银行资本耗竭脆弱性的非线性影响。” 2912, 欧洲中央银行. https://ideas.repec.org/p/ecb/ecbwps/20242912.html.

De Jonghe, O., H. Dewachter, and S. Ongena. 2020. “Bank Capital (Requirements) and Credit Supply: Evidence From Pillar 2 Decisions.” Journal of Corporate Finance 60: S0929119918307521. https://EconP apers.repec.org/RePEc:eee:corfin:v:60:y:2020:i:c:s0929119918307521.
De Jonghe, O., H. Dewachter, 和 S. Ongena. 2020. “银行资本（要求）与信贷供应：来自第二支柱决策的证据。” 企业财务杂志 60: S0929119918307521. https://EconP apers.repec.org/RePEc:eee:corfin:v:60:y:2020:i:c:s0929119918307521.

DelGiovane, P., A. Nobili, and F. M. Signoretti. 2017. “Assessing the Sources of Credit Supply Tightening: Was the Sovereign Debt Crisis Different From Lehman?” International Journal of Central Banking 13, no. 2: 197-234.
DelGiovane, P., A. Nobili, 和 F. M. Signoretti. 2017. “评估信贷供应紧缩的来源：主权债务危机与雷曼兄弟危机有何不同？”《国际中央银行学杂志》13, no. 2: 197-234.
DeMol, C., D. Giannone, and L. Reichlin. 2008. “Forecasting Using a Large Number of Predictors: Is Bayesian Shrinkage a Valid Alternative to Principal Components?” Journal of Econometrics 146, no. 2: 318-328.
DeMol, C., D. Giannone, 和 L. Reichlin. 2008. “使用大量预测变量进行预测：贝叶斯收缩是否是主成分的有效替代？” 计量经济学杂志 146, no. 2: 318-328.

DeNicolò, G. 2015. “Revisiting the Impact of Bank Capital Requirements on Lending and Real Activity.” IMF and CESifo, mimeo.
DeNicolò, G. 2015. “重新审视银行资本要求对贷款和实际活动的影响。”国际货币基金组织和 CESifo，未出版稿。
Doan, T., R. B. Litterman, and C. A. Sims. 1986. “Forecasting and Conditional Projection Using Realistic Prior Distribution.” Staff Report 93, Federal Reserve Bank of Minneapolis.
Doan, T., R. B. Litterman, 和 C. A. Sims. 1986. “使用现实的先验分布进行预测和条件投影。” 斯塔夫报告 93, 明尼阿波利斯联邦储备银行。

Eickmeier, S., B. Kolb, and E. Prieto. 2018. “The Macroeconomic Effects of Bank Capital Requirement Tightenings: Evidence From a Narrative Approach.” 2018-42, Centre for Applied Macroeconomic Analysis, Crawford School of Public Policy, The Australian National University.
Eickmeier, S., B. Kolb, 和 E. Prieto. 2018. “银行资本要求收紧的宏观经济影响：来自叙述方法的证据。” 2018-42，应用宏观经济分析中心，克劳福德公共政策学院，澳大利亚国立大学。
Favara, G., and P. Giordani. 2009. “Reconsidering the Role of Money for Output, Prices and Interest Rates.” Journal of Monetary Economics 56, no. 3: 419-430.
Favara, G., 和 P. Giordani. 2009. “重新考虑货币对产出、价格和利率的作用。”《货币经济学杂志》56, no. 3: 419-430.

Galardo, M., and V. Vacca. 2022. “Higher Capital Requirements and Credit Supply: Evidence From Italy.” 1372, Bank of Italy, Economic Research and International Relations Area. https://ideas.repec.org/p/ bdi/wptemi/td_1372_22.html.
Galardo, M., 和 V. Vacca. 2022. “更高的资本要求与信贷供应：来自意大利的证据。” 1372, 意大利银行，经济研究与国际关系领域。 https://ideas.repec.org/p/ bdi/wptemi/td_1372_22.html.
Gambetti, L., and A. Musso. 2017. “Loan Supply Shocks and the Business Cycle.” Journal of Applied Econometrics 32, no. 4: 764-782.
Gambetti, L., 和 A. Musso. 2017. “贷款供应冲击与商业周期。” 应用计量经济学杂志 32, no. 4: 764-782.
Giannone, D., M. Lenza, D. Momferatou, and L. Onorante. 2014. “Short-Term Inflation Projections: A Bayesian Vector Autoregressive Approach.” International Journal of Forecasting 30, no. 3: 635-644.
Giannone, D., M. Lenza, D. Momferatou, 和 L. Onorante. 2014. “短期通胀预测：一种贝叶斯向量自回归方法。” 国际预测杂志 30, no. 3: 635-644.
Giannone, D., M. Lenza, H. Pill, and L. Reichlin. 2012. “The ECB and the Interbank Market.” Economic Journal 122, no. 564: 467-486.
Giannone, D., M. Lenza, H. Pill, 和 L. Reichlin. 2012. “欧洲央行与银行间市场。” 经济学期刊 122, no. 564: 467-486.

Giannone, D., M. Lenza, and G. E. Primiceri. 2015. “Prior Selection for Vector Autoregressions.” Review of Economics and Statistics 97, no. 2: 436-451.
Giannone, D., M. Lenza, 和 G. E. Primiceri. 2015. “向量自回归的先验选择。” 经济学与统计评论 97, no. 2: 436-451.

Giannone, D., M. Lenza, and L. Reichlin. 2019. “Money, Credit, Monetary Policy and the Business Cycle in the Euro Area.” 885, Federal Reserve Bank of New York.
Giannone, D., M. Lenza, 和 L. Reichlin. 2019. “货币、信贷、货币政策与欧元区的商业周期。” 885, 纽约联邦储备银行。
Gilchrist, S., R. Schoenle, J. Sim, and E. Zakrajšek. 2017. “Inflation Dynamics During the Financial Crisis.” American Economic Review 107, no. 3: 785-823. https://ideas.repec.org/a/aea/aecrev/v107y2017i 3p785-823.html.
Gilchrist, S., R. Schoenle, J. Sim, 和 E. Zakrajšek. 2017. “金融危机期间的通货膨胀动态。” 美国经济评论 107, no. 3: 785-823. https://ideas.repec.org/a/aea/aecrev/v107y2017i 3p785-823.html.

Gropp, R., T. Mosk, S. Ongena, and C. Wix. 2019. “Banks Response to Higher Capital Requirements: Evidence From a Quasi-Natural Experiment.” Review of Financial Studies 32, no. 1: 266-299. https:// ideas.repec.org/a/oup/rfinst/v32y2019i1p266-299.html.
Gropp, R., T. Mosk, S. Ongena, 和 C. Wix. 2019. “银行对更高资本要求的反应：来自准自然实验的证据。” 《金融研究评论》 32, no. 1: 266-299. https:// ideas.repec.org/a/oup/rfinst/v32y2019i1p266-299.html.
Hancock, D., and J. A. Wilcox. 1994. “Bank Capital and the Credit Crunch: The Roles of Risk-Weighted and Unweighted Capital Regulations.” Real Estate Economics 22, no. 1: 59-94.
汉考克，D.，和 J. A. 威尔科克斯。1994。“银行资本与信贷紧缩：风险加权和非加权资本监管的作用。”房地产经济学 22，第 1 期：59-94。

Jardet, C., A. Monfort, and F. Pegoraro. 2013. “New Information Response Functions and Applications to Monetary Policy.” Banque de France, mimeo.
Jardet, C., A. Monfort, 和 F. Pegoraro. 2013. “新信息响应函数及其在货币政策中的应用。”法国银行，未出版稿。

Jiménez, G., S. Ongena, J. L. Peydró, and J. Saurina. 2017. "Macroprudential Policy, Countercyclical Bank Capital Buffers and
吉门内斯，G.，S. 翁根纳，J. L. 佩德罗，和 J. 索里纳。2017。“宏观审慎政策、逆周期银行资本缓冲和

Credit Supply: Evidence From the Spanish Dynamic Provisioning Experiments." Journal of Political Economy 125, no. 6: 2126-2177.
信用供应：来自西班牙动态准备金实验的证据。《政治经济学杂志》125 卷，第 6 期：2126-2177。
Jorda, O. 2005. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review 95, no. 1: 161-182.
Jorda, O. 2005. “通过局部投影估计和推断冲击响应。” 美国经济评论 95, no. 1: 161-182.
Kanngiesser, D., M. Reiner, M. Laurent, and D. Moccero. 2020. “The Macroeconomic Impact of Shocks to Bank Capital Buffers in the Euro Area.” B.E. Journal of Macroeconomics 20, no. 1: 1-17.
Kanngiesser, D., M. Reiner, M. Laurent, 和 D. Moccero. 2020. “欧元区银行资本缓冲冲击的宏观经济影响。” B.E. Journal of Macroeconomics 20, no. 1: 1-17.

Kapetanios, G., H. Mumtaz, I. Stevens, and K. Theodoridis. 2012. “Assessing the Economy-Wide Effects of Quantitative Easing.” Economic Journal 122, no. 564: F316-F347.
Kapetanios, G., H. Mumtaz, I. Stevens, 和 K. Theodoridis. 2012. “评估量化宽松的经济整体影响。” 经济学期刊 122, no. 564: F316-F347.

Koop, G., and D. Korobilis. 2013. “Large Time-Varying Parameter VARs.” Journal of Econometrics 177, no. 2: 185-198.
Koop, G., 和 D. Korobilis. 2013. “大时间变参数 VARs.” 计量经济学杂志 177, no. 2: 185-198.

Krippner, L. 2013. “Measuring the Stance of Monetary Policy in Zero Lower Bound Environments.” Economics Letters 118, no. 1: 135-138.
Krippner, L. 2013. “在零下限环境中衡量货币政策的立场。”经济学通讯 118, no. 1: 135-138.

Lang, J. H., and D. Menno. 2023. “The State-Dependent Impact of Changes in Bank Capital Requirements.” 2828, European Central Bank. https://ideas.repec.org/p/ecb/ecbwps/20232828.html.
Lang, J. H., 和 D. Menno. 2023. “银行资本要求变化的状态依赖影响。” 2828, 欧洲中央银行. https://ideas.repec.org/p/ecb/ecbwps/20232828.html.

Lenza, M., H. Pill, and L. Reichlin. 2010. “Monetary Policy in Exceptional Times.” Economic Policy 25, no. 62: 295-339.
Lenza, M., H. Pill, 和 L. Reichlin. 2010. “特殊时期的货币政策。” 经济政策 25, no. 62: 295-339.

Meeks, R. 2017. “Capital Regulation and the Macroeconomy: Empirical Evidence and Macroprudential Policy.” European Economic Review 95: 125-141.
Meeks, R. 2017. “资本监管与宏观经济：实证证据与宏观审慎政策。” 欧洲经济评论 95: 125-141.
Mendicino, C., K. Nikolov, J. Suarez, and D. Supera. 2020. “Bank Capital in the Short and in the Long Run.” Journal of Monetary Economics 115: 64-79. http://www.sciencedirect.com/science/article/pii/S030439321 9301163.
Mendicino, C., K. Nikolov, J. Suarez, 和 D. Supera. 2020. “短期与长期的银行资本。”《货币经济学杂志》115: 64-79. http://www.sciencedirect.com/science/article/pii/S030439321 9301163.

Mésonnier, J.-S., and A. Monks. 2015. “Did the EBA Capital Exercise Cause a Credit Crunch in the Euro Area?” International Journal of Central Banking 11, no. 3: 75-117.
Mésonnier, J.-S., 和 A. Monks. 2015. “EBA 资本测试是否导致了欧元区的信贷紧缩？” 国际中央银行学杂志 11, no. 3: 75-117.

Mésonnier, J.-S., and D. Stevanovic. 2017. “The Macroeconomic Effects of Shocks to Large Banks’ Capital.” Oxford Bulletin of Economics and Statistics 79, no. 4: 546-569.
Mésonnier, J.-S., 和 D. Stevanovic. 2017. “大型银行资本冲击的宏观经济影响。” 牛津经济与统计公报 79, no. 4: 546-569.
Noss, J., and P. Toffano. 2016. “Estimating the Impact of Changes in Aggregate Bank Capital Requirements on Lending and Growth During an Upswing.” Journal of Banking & Finance 62, no. C: 15-27.
Noss, J., 和 P. Toffano. 2016. “估计在经济上行期间，银行资本要求变化对贷款和增长的影响。” 《银行与金融杂志》 62, no. C: 15-27.
Prieto, E., S. Eickmeier, and M. Marcellino. 2016. “Time Variation in Macro Financial Linkages.” Journal of Applied Econometrics 31, no. 7: 1215-1233.
Prieto, E., S. Eickmeier, 和 M. Marcellino. 2016. “宏观金融联系的时间变化。” 应用计量经济学杂志 31, no. 7: 1215-1233.

Primiceri, G. E. 2005. “Time Varying Structural Vector Autoregressions and Monetary Policy.” Review of Economic Studies 72, no. 3: 821-852.
Primiceri, G. E. 2005. “时间变化的结构向量自回归与货币政策。”经济研究评论 72, no. 3: 821-852.
Raja, A. 2023. “The Impact of Changes in Bank Capital Requirements.” 1004, Bank of England. https://ideas.repec.org/p/boe/boeewp/ 1004.html.
Raja, A. 2023. “银行资本要求变化的影响。” 1004, 英国银行。 https://ideas.repec.org/p/boe/boeewp/ 1004.html.

Romer, C. D., and D. H. Romer. 2004. “A New Measure of Monetary Shocks: Derivation and Implications.” American Economic Review 94, no. 4: 1055-1084.
Romer, C. D., 和 D. H. Romer. 2004. “货币冲击的新测量：推导与影响。” 美国经济评论 94, no. 4: 1055-1084.

Romer, C. D., and D. H. Romer. 2017. “New Evidence on the Aftermath of Financial Crises in Advanced Countries.” American Economic Review 107, no. 10: 3072-3118.
Romer, C. D., 和 D. H. Romer. 2017. “发达国家金融危机后果的新证据。” 美国经济评论 107, no. 10: 3072-3118.
Rostagno, M., C. Altavilla, G. Carboni, et al. 2019. “A Tale of Two Decades: The ECB’s Monetary Policy at 20.” 2346, European Central Bank. https://ideas.repec.org/p/ecb/ecbwps/20192346.html.
Rostagno, M., C. Altavilla, G. Carboni, et al. 2019. “两个十年的故事：欧洲央行的货币政策 20 年。” 2346, 欧洲中央银行. https://ideas.repec.org/p/ecb/ecbwps/20192346.html.
Sims, C. A., and T. Zha. 1998. “Bayesian Methods for Dynamic Multivariate Models.” International Economic Review 39, no. 4: 949-968.
Sims, C. A., 和 T. Zha. 1998. “动态多元模型的贝叶斯方法。” 国际经济评论 39, no. 4: 949-968.
von Borstel, J., S. Eickmeier, and L. Krippner. 2016. “The Interest Rate Pass-Through in the Euro Area During the Sovereign Debt Crisis.” Journal of International Money and Finance 68: 386-402.
von Borstel, J., S. Eickmeier, 和 L. Krippner. 2016. “欧元区主权债务危机期间的利率传导。” 国际货币与金融杂志 68: 386-402.

Waggoner, D. F., and T. Zha. 1999. “Conditional Forecasts In Dynamic Multivariate Models.” Review of Economics and Statistics 81, no. 4: 639-651.
Waggoner, D. F., 和 T. Zha. 1999. “动态多元模型中的条件预测。” 经济学与统计评论 81, no. 4: 639-651.

Wu, J. C., and F. D. Xia. 2016. “Measuring the Macroeconomic Impact of Monetary Policy at the Zero Lower Bound.” Journal of Money, Credit and Banking 48, no. 2-3: 253-291.
吴, J. C., 和 F. D. 夏. 2016. “在零下限下衡量货币政策的宏观经济影响。”《货币、信贷与银行学报》48, no. 2-3: 253-291.

Supporting Information支持信息

Additional supporting information can be found online in the Supporting Information section.
额外的支持信息可以在在线支持信息部分找到。

RESEARCH ARTICLE研究文章

Bank Capital Requirements, Lending Supply, and Economic Activity: A Scenario Analysis Perspective 银行资本要求、贷款供应与经济活动：情景分析视角

Abstract摘要

1 | Introduction1 | 引言

2 | Related Literature2 | 相关文献

3 | Institutional Background and the Evolution of Tier 1 Capital Ratio3 | 机构背景与一级资本充足率的演变

Tier 1 capital ratio一级资本充足率

4 | Empirical Framework4 | 实证框架

4.1 | Fixed-Coefficients Bayesian VAR Model4.1 | 固定系数贝叶斯 VAR 模型

4.2 | Time-Varying Coefficients Bayesian VAR Model4.2 | 时变系数贝叶斯 VAR 模型

4.3 | The Data4.3 | 数据

4.4 | How Are Data Selected?4.4 | 数据是如何选择的？

4.5 | Conditional Forecasting and Scenario Analysis4.5 | 条件预测与情景分析

5 | Recovering the Impact of Regulatory and Supervisory Actions5 | 恢复监管和监督行动的影响

5.1 | Step 1: Estimating the "Size" of the Policy Actions5.1 | 第一步：估算政策行动的“规模”

5.2 | Step 2: Estimating the Impact of the Policy Actions on Lending Supply and Economic Activity5.2 | 第 2 步：估计政策措施对贷款供应和经济活动的影响

5.3 | Discussion5.3 | 讨论

6 | Baseline Results6 | 基准结果

7 | Increases in Bank Capital Requirements or Structural Breaks?7 | 银行资本要求的增加或结构性断裂？

7.1 | Time-Varying Coefficients and Volatility7.1 | 时间变动系数与波动性

7.2 | "In-Sample" Counterfactuals7.2 | “样本内”反事实

Abstract摘要

8 | Additional Analyses and Robustness8 | 附加分析与稳健性

8.1 | Evidence From the Euro Area BLS8.1 | 来自欧元区 BLS 的证据

8.2 | Forecasts of Main Macroeconomic and Credit Variables Under the Policy Scenario8.2 | 政策情景下主要宏观经济和信贷变量的预测

8.3 | Including the US Tier 1 Ratio as Global Regulation Factor8.3 | 包括美国一级资本充足率作为全球监管因素

8.4 | A "Placebo" Test8.4 | 一项“安慰剂”测试

8.5 | Separating Tier 1 Capital and RWAs8.5 | 分离一级资本和风险加权资产

9 | Concluding Remarks9 | 结论

Acknowledgments致谢

Data Availability Statement数据可用性声明

Endnotes尾注

References参考文献

Supporting Information支持信息

Bank Capital Requirements, Lending Supply, and Economic Activity: A Scenario Analysis Perspective
银行资本要求、贷款供应与经济活动：情景分析视角

2 | Related Literature
2 | 相关文献

3 | Institutional Background and the Evolution of Tier 1 Capital Ratio
3 | 机构背景与一级资本充足率的演变

Tier 1 capital ratio
一级资本充足率

4 | Empirical Framework
4 | 实证框架

4.1 | Fixed-Coefficients Bayesian VAR Model
4.1 | 固定系数贝叶斯 VAR 模型

4.2 | Time-Varying Coefficients Bayesian VAR Model
4.2 | 时变系数贝叶斯 VAR 模型

4.3 | The Data
4.3 | 数据

4.4 | How Are Data Selected?
4.4 | 数据是如何选择的？

4.5 | Conditional Forecasting and Scenario Analysis
4.5 | 条件预测与情景分析

5 | Recovering the Impact of Regulatory and Supervisory Actions
5 | 恢复监管和监督行动的影响

5.1 | Step 1: Estimating the "Size" of the Policy Actions
5.1 | 第一步：估算政策行动的“规模”

5.2 | Step 2: Estimating the Impact of the Policy Actions on Lending Supply and Economic Activity
5.2 | 第 2 步：估计政策措施对贷款供应和经济活动的影响

6 | Baseline Results
6 | 基准结果

7 | Increases in Bank Capital Requirements or Structural Breaks?
7 | 银行资本要求的增加或结构性断裂？

7.1 | Time-Varying Coefficients and Volatility
7.1 | 时间变动系数与波动性

7.2 | "In-Sample" Counterfactuals
7.2 | “样本内”反事实

8 | Additional Analyses and Robustness
8 | 附加分析与稳健性

8.1 | Evidence From the Euro Area BLS
8.1 | 来自欧元区 BLS 的证据

8.2 | Forecasts of Main Macroeconomic and Credit Variables Under the Policy Scenario
8.2 | 政策情景下主要宏观经济和信贷变量的预测

8.3 | Including the US Tier 1 Ratio as Global Regulation Factor
8.3 | 包括美国一级资本充足率作为全球监管因素

8.4 | A "Placebo" Test
8.4 | 一项“安慰剂”测试

8.5 | Separating Tier 1 Capital and RWAs
8.5 | 分离一级资本和风险加权资产

9 | Concluding Remarks
9 | 结论

Data Availability Statement
数据可用性声明