核心问题
当通胀或增长发生变化时,央行实际能推动什么、通过哪个渠道、需要多长时间?
- 政策分析
- 研究
- 经济史
货币政策是最显眼的宏观工具,因为它行动快,决策日期也清楚。但这种可见性容易误导:公告是即时的,经济影响却分阶段到来。
利率决定首先改变短期融资价格。之后,它会通过信贷市场、汇率、资产估值和对未来经济路径的预期继续传导。
政策目标
当通胀或增长发生变化时,央行实际能推动什么、通过哪个渠道、需要多长时间?
利率、资产负债表政策、流动性工具和前瞻指引,都是先影响金融条件,再逐步影响招聘、定价和支出。
货币政策最快影响资产价格和融资条件,随后才更慢地影响实体经济中的借贷、招聘和支出决策。
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。
政策利率
政策利率确定短端定价环境,并影响银行、市场和借款人如何重新评估资金成本。
资产负债表政策
资产购买、缩表和抵押品操作,试图在政策利率不足时影响长期收益率、市场运行和金融条件。
前瞻指引
央行通过沟通影响未来利率、通胀和衰退风险的预期,早于完整现金流影响出现之前。
流动性工具
当问题是市场失灵或融资压力,而不是普通周期性过热时,紧急工具和常设支持就很重要。
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。
预期与可信度
利率决策在很大程度上通过人们对央行下一步行动的预期发挥作用。如果居民和企业相信通胀会回到2%,工资和价格的设定就会完成一部分工作,政策利率需要做的就更少。沃尔克时期展示了相反的情形:由于市场预期高通胀会卷土重来,长期收益率直到1982年仍居高不下。
信贷与资产负债表
政策利率不是直接作用于经济,而是通过银行和市场传导。浮动利率贷款会立即重新定价,固定利率债务则要等到再融资时才调整,因此同一次加息对一些借款人很早见效,对另一些借款人则要数年之后。贷款机构自身融资成本上升时,也会收紧放贷标准。
汇率与外部溢出
国内利率上升会吸引资本流入,往往推高本币,从而压低进口价格、抬高出口价格。由于许多外国企业和政府以美元借款,美联储收紧政策也会收紧海外的金融条件,外币债务规模大的地方受影响最深。
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 repricing
Minutes to days. Yield curves, exchange rates, equity prices, and credit spreads react before the real economy does.
Credit and spending
Months to quarters. Refinancing, loan demand, hiring, and investment update as existing contracts reset.
Inflation pass-through
Often one to two years. Services inflation, wage setting, and expectations move slowly unless credibility breaks.
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。
Mandate and reaction function
Central banks must explain how inflation, employment, financial stability, and credibility enter the rule they are following.
Effective lower bound
Once short rates approach the lower bound, balance-sheet tools and guidance carry more of the transmission burden.
Fiscal and financial plumbing
Rate policy lands through banks, markets, Treasury issuance, and regulated balance sheets rather than through a single direct lever.
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。
Mistaking relative prices for persistent inflation
A supply shock can look like demand overheating. Tightening into the wrong diagnosis can cut output without fixing the source of price pressure.
Moving too late
If households and firms stop believing the target, the policy rate must do more damage to rebuild credibility.
Breaking a fragile balance sheet
A rate path that looks mild in aggregate data can be severe for banks, leveraged firms, floating-rate borrowers, and sovereigns with short maturities.
取舍
任何有用的干预都会牵动别的东西。
迟一步行动,还是依据预测行动
利率调整的效果有滞后,因此委员会必须依据预测而不是已经看到的情况行事。等到证据确凿再动手,追赶就会很急促:美联储自2022年3月起在16个月内把利率提高了5.25个百分点。依据一个后来证明是错误的预测行动,则会白白打断一次复苏。
公众可检验的规则,还是相机判断的空间
像泰勒规则这样的规则让政策可预测,也让外部人士能够检验。但规则需要一个中性利率和一个潜在产出水平,而这两者都无人能直接观测,这正是泰勒本人把这类公式称为参照标准的原因。联邦公开市场委员会(FOMC)定期参考多条规则,但不机械地遵循其中任何一条。
全经济一个利率,成本却不均
中央银行为所有地区和行业设定同一个利率。成本首先落在借款人以及建筑业、制造业等对利率敏感的工作上。在1981-82年的衰退中,商品生产部门约占就业的30%,却承担了1982年90%的岗位流失。
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。
- Policy pathPolicy rate seriesTrack the target rate, expected path, and the distance between market pricing and central-bank guidance.
- Inflation expectationsInflation expectationsCompare survey and market-implied expectations before treating realized inflation as the full signal.
- Financial conditionsFinancial conditionsWatch credit spreads, real yields, equity conditions, and dollar pressure for signs that transmission is tightening faster than the policy rate.
历史案例
检验这项政策诊断的真实案例。
1979-1983
The Volcker disinflation
A major case of credibility, lag, and pain: inflation came down, but only through a recession that showed how unevenly monetary policy lands.
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.
2022-2024
Fast hikes after the reopening inflation shock
Central banks raised rates into supply repair, labor-market strength, and balance-sheet sensitivity. The case is a central modern test of lag uncertainty.
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。
Taylor rule
Taylor (1993). The benchmark reaction function: nominal policy rate equals the neutral real rate plus current inflation, with equal weights on the inflation and output gaps.
Fisher identity
The nominal rate is the real rate plus expected inflation. Why a credible inflation target lowers nominal rates without lowering real rates.
Expectations-augmented Phillips curve
New Keynesian form. The slope coefficient \(\kappa\) is the part economists argue about: low \(\kappa\) gives small inflation cost from output overshoots, high \(\kappa\) the opposite.
What the Taylor rule actually says
John Taylor's 1993 Carnegie-Rochester paper [1] proposed a simple rule and compared it with the actual federal funds rate over 1987-1992: start from an equilibrium real rate of 2 percent, add inflation over the previous four quarters, add half a point for every point inflation runs above a 2 percent target, and add half a point for every percent real GDP runs above trend. Taylor wrote that the rule fit actual policy in those years remarkably well. It began as a description. It is now a standard benchmark: the FOMC regularly consults the prescriptions of several such rules alongside its other information [2].
The coefficients are secondary. The structure is what matters: it forces a policymaker to declare two unobservable numbers before announcing a rate. The neutral real rate \(r^*\) is the level at which monetary policy is neither tight nor loose. Potential output \(y^*\) is the level the economy can sustain without generating inflation pressure. Both move slowly. Both are estimated, not observed. Alan Greenspan made the point at Stanford on September 5, 1997: rules of this kind are at best "guideposts", because they depend on the equilibrium real federal funds rate and the economy's production potential, and those are inferred from past behavior [3]. Either can be wrong by enough to reverse the rule's verdict on whether policy is tight or loose.
The years before the pandemic show how much rides on \(r^*\). In Taylor's own example, with inflation and output both on target, the rule returns a 4 percent funds rate: a 2 percent real rate plus 2 percent inflation. FOMC participants no longer believed that number. The median projection of the longer-run federal funds rate fell from 4.25 percent in January 2012 [4] to 2.5 percent in June 2019 [5], which implies a neutral real rate near 0.5 percent. Run with that input, the same rule and the same on-target economy return 2.5 percent, the top of the range the Fed reached in December 2018 [6]. The Fed held there in January 2019 and by the end of October had cut the range by three quarters of a point [7]. The August 2020 revision of its strategy statement put the shift in writing: the federal funds rate consistent with the Committee's goals over the longer run "has declined relative to its historical average" [8]. The rule had not changed. One of its inputs had.
Why the Fisher decomposition matters for breaking inflation
The Fisher identity says \(i = r + \mathbb{E}[\pi]\). It is an accounting identity, not a behavioral claim. It tells you nothing about which side moves when the central bank announces a hike. The empirical question is whether a 100 basis-point increase in the policy rate moves \(r\) by 100 basis points (the rule of thumb in the textbook), or whether part of it bleeds into expected inflation through the credibility channel.
Volcker's disinflation is the standard case for studying that channel. He became chairman in August 1979 with inflation running above 11 percent [9]. Goodfriend and King date the deliberate disinflation from late 1980, when the federal funds rate rose to 19 percent [10]. The ten-year Treasury yield did not follow inflation down: while inflation fell from over 10 percent in early 1981 to under 6 percent by mid-1982, the ten-year rate rose from around 13 percent to over 14 percent. They read that as markets expecting high inflation to return. Credibility came late. Long-term rates began to decline only in late 1982, by which time unemployment had reached nearly 11 percent [11]. The episode shows what a rate increase costs when the public does not yet believe the central bank.
The 2022 tightening is the contrast. When the Fed raised its target range on March 16, 2022 [12], short-run expectations had already moved: the New York Fed's survey put median expected inflation one year ahead at 6.0 percent in February 2022 [13]. Longer-term measures had moved up less [14]. The target range rose from 0 to 0.25 percent to 5.25 to 5.5 percent by July 2023 [15], an increase of 5.25 percentage points in 16 months. Chair Powell later judged that this, together with the unwinding of pandemic supply disruptions, brought inflation much closer to target without the painful rise in unemployment of earlier disinflations [16]. The difference from 1981 is the behavior of longer-term expectations, which moved far less than short-term ones.
How flat is the Phillips curve
Olivier Blanchard's January 2016 policy brief for the Peterson Institute, "The US Phillips Curve: Back to the 60s?", reached four conclusions [17]. Low unemployment still pushes inflation up. Expectations had become anchored, so unemployment was related to the level of inflation and no longer to its change, as in the 1960s. The slope had declined substantially, and the decline dated from the 1980s, not from the 2008 crisis. And the relation was noisy. Janet Yellen drew the policy conclusion in her speech of September 26, 2017. She accepted that the influence of labor utilization on inflation had become "quite modest", but warned that without further modest rate increases over time the labor market could eventually become overheated, and that it would be imprudent to keep policy on hold until inflation was back to 2 percent [18]. A flat curve cuts both ways. It makes running the economy hot look cheap, and it makes an inflation problem expensive to undo. Both readings turn on \(\kappa\).
The slope is hard to identify because the demand shocks are not random. A central bank that targets inflation suppresses the variation in inflation that comes from demand shocks, leaving mostly cost shocks in the residual. Cost shocks move output and inflation in opposite directions; demand shocks move them together. If you regress observed inflation on observed unemployment in a credible regime, the slope you recover is biased toward zero. The curve looks flatter than it is because the regime has been doing its job.
Hazell, Herreño, Nakamura, and Steinsson (2022) [19] estimate that the Phillips-curve slope was small even in the early 1980s and declined only modestly afterward. They attribute much of the apparent historical steepness and later inflation stability to shifts in long-run inflation expectations. The result cautions against treating a raw time-series correlation as a stable structural slope.
How the Fed's operating framework works (as of November 2026)
The federal funds market is now small. A Federal Reserve staff paper documents average daily trading volume of $72 billion when the effective funds rate sat at or below the interest rate on reserves and $67 billion when it sat above, during April 2018 to February 2020, with Federal Home Loan Banks, which cannot earn interest on reserves, providing almost all of the lending [20]. The market that matters for transmission is overnight repo. A 2025 FEDS Note estimates the gross size of the U.S. repo market at $11.9 trillion in 2024 [21]. The Fed's policy rate is still defined as a target range for the federal funds rate. What keeps the funds rate inside that range is a set of rates the Fed sets directly.
Since January 2019 the FOMC has said it will run an ample-reserves regime, in which control over short-term rates comes primarily from the Federal Reserve's administered rates and active management of the supply of reserves is not required [22]. Three rates do the work [23]. Interest on reserve balances (IORB), set by the Board, is what banks earn on deposits at the Fed [24]. The overnight reverse repo (ON RRP) rate, a supplementary tool, does the same job for a wider set of counterparties: none of them should lend overnight for less than the Fed will pay [25]. Standing repo operations lend against Treasury and agency securities and help put a ceiling on overnight rates [26]. After the meeting of October 28, 2026 the settings were a target range of 3.75 to 4 percent, IORB at 3.90 percent, ON RRP at 3.75 percent, and standing repo at 4.0 percent [27].
The balance sheet supplies the reserves that make this work. The Fed let its securities holdings run off from June 2022 and ended the runoff on December 1, 2025, after they had fallen by more than $2.2 trillion [28]. On December 10, 2025 the Committee judged that reserve balances had declined to ample levels and said it would buy shorter-term Treasury securities as needed to keep them there [29]. The caution has a history. In mid-September 2019, with reserves declining, the secured overnight rate jumped above 5 percent and the federal funds rate moved above its target range [30].
Communication is the other instrument. Every scheduled meeting now produces a statement, an implementation note, and a press conference, and four meetings a year add the Summary of Economic Projections [31]. Because a rate decision works through expectations of future rates, what the Committee says about the path matters as well as what it does. The credibility of that communication is a policy variable in its own right.
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。