核心问题
如果金融体系本身就是约束,哪些工具能保持信贷渠道运转,又不把每个私人错误都社会化?
金融稳定政策位于宏观经济学、银行监管和危机管理的交界处。
因此,这些工具常常不同于常规货币或财政措施。目标不是简单刺激需求或压低通胀,而是防止金融体系把坏冲击放大为全面崩塌。
政策目标
如果金融体系本身就是约束,哪些工具能保持信贷渠道运转,又不把每个私人错误都社会化?
当问题不只是需求疲弱,而是信贷渠道受损或系统性脆弱时,银行资本规则、紧急贷款、流动性支持和监管行动就很关键。
金融稳定政策较少由某个头条利率衡量,更多取决于它能否防止融资、支付和信贷渠道放大冲击。
工具与手段
政策真正可以调动的杠杆。每种工具都有各自的操作目标、信息需求和政治成本。
资本与流动性规则
监管标准旨在确保核心金融中介能在恐慌扩散前吸收损失并应对融资压力。
紧急贷款
最后贷款人工具支持有偿付能力、但面临暂时流动性压力的机构,并抑制抛售循环。
市场支持
对核心市场或融资渠道的支持,可以在私人资产负债表被迫同时去杠杆时防止定价失灵。
监管干预
限制、再资本化、处置规划和监管压力,在问题集中于具体机构内部时尤其重要。
传导渠道
政策工具影响价格、产出、就业、信贷、对外收支或产能的渠道。
银行资本与信贷供给
损失会侵蚀银行资本,资本不足的银行可以通过减少放贷来恢复资本比率。金融稳定政策作用于这一环节:资本规则、流动性贷款和资本重组,使贷款机构在确认损失的同时仍能提供信贷。美联储工作人员在2013年估计,潜在产出比2007年之前的路径低约7%,由此可见这一环节断裂时的代价。
信心与挤兑
挤兑是信念的转变。存款人和短期贷款人之所以撤资,是因为预期别人会撤资,所以一项可信的担保无需投入很多资金就能终结恐慌。2023年3月12日,也就是硅谷银行被关闭两天后,监管机构宣布其所有存款人都将得到全额保护。
跨境融资
大型银行在多个国家以多种货币融资,而监管、存款保险和财政支持仍以国家为单位。因此,压力在融资市场上的传播速度快于任何单一当局的反应速度,这正是美联储在2007-2009年危机期间与外国央行建立货币互换额度的原因。
时滞
效果究竟何时显现。有些价格会立即调整;数量、合同、招聘和投资通常反应得更晚。
Market stress
Minutes to days. Funding spreads, deposit flows, and collateral haircuts can change faster than supervisors can publish a report.
Supervisory repair
Weeks to quarters. Examinations, recapitalization, asset sales, and resolution planning take time.
Credit recovery
Quarters to years. Even after panic fades, banks may rebuild capital by shrinking risk rather than expanding credit.
制度约束
决定一项技术上自洽的政策能否实际使用的法律、政治、资产负债表和操作层面的限制。
Moral hazard
Backstops can stop panic while teaching markets to expect future rescues. Design must separate liquidity repair from investor protection.
Resolution authority
A credible regime needs legal power to close, merge, recapitalize, or wind down failing institutions.
Cross-border balance sheets
Funding markets and bank groups are international, but deposit insurance, supervision, and fiscal backstops remain national.
失灵方式
政策可能失灵、来得太晚或作用于错误边际的各种方式。
Liquidity support for insolvency
A facility can buy time, but it cannot make a bad asset book sound without capital or loss recognition.
Backstop ambiguity
If markets cannot tell who is protected and who takes losses, the announcement can fuel rather than stop panic.
Regulatory blind spots
Risk often migrates to shadow banking, nonbank finance, or off-balance-sheet structures when rules bind only the visible perimeter.
取舍
任何有用的干预都会牵动别的东西。
救一家企业,还是救整个市场
救助单一濒临倒闭的企业能止住眼前的损害,却也奖励了冒险的那家企业。面向整个市场开放的支持削弱了这种激励,代价是也会贷款给许多并不需要的企业。自2010年起,美国法律倾向于后者:紧急贷款必须通过具有广泛参与资格的计划发放。
事先声明的规则,还是恐慌时的自由裁量
事先宣布的规则,例如不向资不抵债的企业放贷,只有在人们相信大企业倒下时它依然有效,才能抑制道德风险。相机裁量让当局在意外状况出现时能迅速行动,却也让市场学会期待下一次援手。要求财政部长批准任何紧急计划,正是在保留裁量权的同时为它加上一道制衡。
谁来承担损失
每一次处置都要决定由谁承担损失:股东、债券持有人、未受保存款人、存款保险基金,还是纳税人。保护债权人能平息恐慌,但会把成本转嫁给公众和缴纳保费的银行。让债权人承担损失更公平,却可能把挤兑蔓延到下一家机构。
需要监测的数据
关闭本页之后,继续追踪那些可能推翻这一诊断的数据序列。
- Credit spreadsCredit-spread indicatorsSpreads show whether risk appetite, funding pressure, or default expectations are tightening the credit channel.
- Bank lendingBank lending seriesLoan growth, standards, and delinquency rates show whether the banking system is transmitting or absorbing stress.
- Funding conditionsFinancial conditionsFunding spreads, reserves, deposits, and market liquidity show whether stress has become systemic.
历史案例
检验这项政策诊断的真实案例。
2008-2014
Zero rates, QE, and emergency facilities
When the usual rate tool hit its floor, macro policy widened into balance-sheet policy, backstops, and a much louder debate over what central banks can really do alone.
2010-2012
Euro-area bank-sovereign loop
Bank balance sheets, sovereign debt, and institutional design reinforced each other until backstops and supervision changed the regime.
2023
Deposit flight and duration risk
Fast rate hikes revealed how uninsured deposits, securities losses, and confidence can turn an accounting problem into a liquidity event.
关键恒等式
本页其余部分所依据的方程。
把恒等式当作检验政策逻辑的工具,而不是指望一个方程就能决定整个政策体制。
Diamond-Dybvig run condition
Diamond and Dybvig (1983). Two equilibria coexist: one where patient depositors wait, one where everyone runs. A bank's demand-deposit promise makes the run equilibrium possible.
Countercyclical capital buffer
Basel III. A stylized version of the buffer guide: it rises with the credit-to-GDP gap \(\hat{c}_t - c^*\) at a rate \(\alpha\). In the Basel Committee's guidance the guide is zero until the gap passes 2 percentage points and reaches its maximum of 2.5 percent of risk-weighted assets at a gap of 10 points. The guide is a reference point, not a rule. Each national authority sets the buffer using its own judgment, announces an increase up to 12 months before it applies, and can release the buffer at once.
Basel leverage ratio
A backstop that ignores risk weights, introduced after 2008. The 3 percent figure is the Basel baseline. Global systemically important banks carry a leverage-ratio buffer on top of it, and national implementations differ. The risk-weighted ratio is usually the constraint that binds first. The leverage ratio binds when a bank holds many assets with low risk weights.
Diamond-Dybvig and the structure of bank runs
Douglas Diamond and Philip Dybvig published "Bank Runs, Deposit Insurance, and Liquidity" in the Journal of Political Economy in 1983 [1]. The model has three dates, two types of depositor, and one bank. At the first date everyone deposits. At the second some depositors learn that they are impatient and need to consume now, while the patient ones can wait until the third. The bank invests in a long-term asset and promises every depositor the option to withdraw early. The model has two equilibria. In one, only impatient depositors withdraw early; the bank pays them and the patient ones wait for the higher long-term return. In the other, everyone withdraws because everyone expects everyone else to, and the bank has to liquidate the long-term asset at a loss.
The point of the model is that the second equilibrium is not irrational. It is a Nash equilibrium given what the bank has promised. The paper shows that government deposit insurance can produce a contract that prevents the run, and its authors shared the 2022 Nobel prize in economics with Ben Bernanke for research on banks and financial crises [2]. The guarantee has a cost: it changes the price of risk, so insured banks have to be supervised.
The 2008 panic and the 2023 failure of Silicon Valley Bank are both Diamond-Dybvig stories with different runners. In 2008 the runners were investors in money-market funds and other short-term lenders, not retail depositors. On September 19, amid concern that the funds' share values would fall below one dollar, the Treasury announced a temporary guarantee for money-market funds [3]. The Federal Reserve lent directly to the markets those funds were selling into: it announced the Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility the same day [4] and the Commercial Paper Funding Facility on October 7 [5]. In 2023 the runners were uninsured depositors. Silicon Valley Bank lost more than $40 billion of deposits on March 9 and was closed the next day [6]. On March 12 the Treasury, the Federal Reserve, and the FDIC invoked the systemic-risk exception so that all depositors were protected [7]. The institutions differed. The structure of the problem was the same.
Basel I to Basel III, in three constraints
Basel I (1988) set a single requirement: capital of at least 8 percent of risk-weighted assets, of which at least half had to be core, or Tier 1, capital [8]. The risk weights were crude. Claims on OECD central governments got a zero weight. Residential mortgages got 50 percent. The accord was approved by the central bank governors of the G10 and was to be in place by the end of 1992; in time it was introduced in virtually all countries with internationally active banks [9].
Basel II (2004) kept a standardised approach and added a second route for credit risk, the internal ratings-based approach, under which a bank's own risk assessments feed the capital calculation once its supervisor is satisfied with its systems [10]. Internal value-at-risk models had already been allowed for market risk by an amendment of 1996. The framework stood on three pillars: minimum capital requirements, supervisory review, and market discipline through disclosure. The aim was capital that tracked risk more closely. The Committee's own later verdict was severe: banks entered the crisis with too much leverage and inadequate liquidity buffers, and at its peak many stakeholders lost faith in reported risk-weighted capital ratios [9].
Basel III (agreed in 2010 and completed in 2017) added constraints that do not depend on risk weights: a leverage ratio, a liquidity coverage ratio meant to provide enough cash to cover funding needs over 30 days of stress, and a net stable funding ratio aimed at maturity mismatch [9]. It also raised the risk-weighted minima, and the tiers are worth keeping distinct. Common Equity Tier 1 must be at least 4.5 percent of risk-weighted assets, Tier 1 at least 6 percent, and total capital at least 8 percent, with a capital conservation buffer of 2.5 percent of CET1 on top [11]. The widely quoted 7 percent figure is the 4.5 percent CET1 minimum plus that buffer; it is a CET1 number, not a Tier 1 minimum. The countercyclical capital buffer, between zero and 2.5 percent, is set at the discretion of each national authority. Global systemically important banks face further requirements, and national implementation varies by jurisdiction.
The 2023 failure of Silicon Valley Bank showed the limits of how the framework was applied in the United States. A 2018 law raised the asset threshold for enhanced prudential standards from $50 billion to $250 billion, and the Federal Reserve's 2019 tailoring rule applied a less stringent set of standards to banks above $100 billion. Silicon Valley Bank's parent had about $212 billion in assets when it failed. Because of transition periods in the rules, supervisory stress testing, the liquidity coverage ratio, and the net stable funding ratio had not yet been applied to it. Its deposits were largely uninsured and invested mainly in longer-term securities that lost value as interest rates rose. The Federal Reserve's own review concluded that its tailoring approach had impeded effective supervision by reducing standards, and estimated that under the rules in force before 2019 the bank would have faced the full liquidity coverage ratio and been about 9 percent short of the liquid assets it required in December 2022 [12].
The 2008 collapse and the policy toolkit it built
The conventional periodization splits the 2007-2009 crisis into three phases. Phase one ran from the summer of 2007 to September 2008. The Federal Reserve responded with new lending facilities tied to its traditional role as lender of last resort: the Term Auction Facility for banks, and the Term Securities Lending Facility and the Primary Dealer Credit Facility for the primary dealers [13].
Phase two was the panic of autumn 2008. On September 6 the Federal Housing Finance Agency placed Fannie Mae and Freddie Mac into conservatorship, with the Treasury committing to keep them solvent [14]. On September 16 the Federal Reserve authorized a loan of up to $85 billion to the insurer AIG under section 13(3) of the Federal Reserve Act [15]. On November 25 the Fed announced its first large-scale purchases: up to $100 billion of debt issued by the housing agencies and up to $500 billion of mortgage-backed securities [16]. On December 16 the FOMC set a target range for the federal funds rate of 0 to 0.25 percent [17].
Phase three was the slow recovery. The Fed did not raise its target range until December 16, 2015 [18]. Federal Reserve staff estimated in 2013 that potential output was about 7 percent below the path it had seemed to be on before 2007, and argued that a significant part of that damage plausibly came from the weakness in demand itself, through hysteresis in the labor market, and not only from a slower trend [19].
The lasting institutional changes were the Dodd-Frank Act in the United States, signed on July 21, 2010 [20], and the Single Supervisory Mechanism in the euro area, under which the European Central Bank took over the supervision of euro-area banks on November 4, 2014 [21]. Dodd-Frank also narrowed the Federal Reserve's emergency-lending power. Section 13(3) had been part of the Federal Reserve Act since 1932. Since 2010 it can be used only for programs with broad-based eligibility, not to keep a single firm out of bankruptcy, borrowers must not be insolvent, and no program may be set up without the prior approval of the Secretary of the Treasury [22]. The Volcker Rule restricted banks from trading for their own account. The Financial Stability Oversight Council was created to watch the financial system as a whole and can designate nonbank firms for Federal Reserve supervision. It designated four in 2013 and 2014, and by October 2018 it had rescinded three of those designations [23].
情景 · 解读政策的方法
在套用政策框架之前,先核对工具、渠道、时滞和失灵方式
工具
在判断政策立场之前,先说清具体使用的是哪种工具。同一个目标可以通过利率、支出、税收、监管、沟通或资产负债表操作来实现。
传导
追踪政策变动如何经由家庭、企业、银行、市场、预期、汇率和公共部门资产负债表传导。
时滞
区分宣布、实施、市场反应、实体经济反应以及数据发布的时间。
失灵方式
说明哪些情况会让政策适得其反、生效太晚、效果外溢到国外、只是转移风险,或无法触及受约束的部门。