The Bank of Japan's Zero Rates: Inventing QE, 1999–2006

On 3 November 1997, Sanyo Securities failed to repay roughly ¥1 billion it had borrowed overnight in Tokyo's uncollateralised call market. Measured against a market that turned over tens of trillions of yen, the sum was nothing, and the default was almost clerical — the firm had filed for court-supervised reorganisation that morning and its accounts were frozen. What mattered was that it had never happened before. Japanese financial institutions lent to one another unsecured and overnight on the understanding that the Ministry of Finance would not allow a counterparty to fail. That understanding had just been tested in public and found to be worth nothing.

Lending in the call market contracted within days. Hokkaido Takushoku Bank, one of the country's city banks, collapsed on 17 November. A week later Yamaichi Securities, founded in 1897 and one of the Big Four brokerages, announced it would wind itself up, carrying some ¥260 billion of losses it had hidden in offshore vehicles for years. Its president, Shohei Nozawa, broke down at the press conference and said what became the most-replayed sentence of the Japanese financial crisis: "The fault is mine. The employees have done nothing wrong. Please, help them find work."

An Independent Central Bank, Cornered on Arrival

Japan's new Bank of Japan Law took effect on 1 April 1998, ending half a century of Ministry of Finance control and giving the Bank statutory independence over monetary policy for the first time since 1942. Masaru Hayami, a former trading-house executive of orthodox instincts, had become governor twelve days earlier. He inherited an institution that was newly free to choose and had almost nothing left to choose from.

Policy rates were already close to the floor. The official discount rate had stood at 0.5 per cent since September 1995, cut there during the long unwinding of the property and equity boom described in the Japanese asset bubble of 1985–1990. On 9 September 1998 the Policy Board lowered its target for the uncollateralised overnight call rate to 0.25 per cent. Six weeks later the Long-Term Credit Bank of Japan was nationalised; Nippon Credit Bank followed in December. Japanese banks borrowing dollars abroad were being charged a premium over their international peers that the market openly called the Japan premium.

Prices were falling, not merely rising slowly. Consumer prices excluding fresh food — the Bank's preferred core measure — turned negative on a year-on-year basis in 1999 and stayed there, with brief exceptions, for the next six years. A falling price level raises real interest rates even when nominal rates do not move, which meant that monetary policy was tightening on its own while the Policy Board sat still.

So on 12 February 1999 the Board did something no major central bank had done in the modern era. It instructed the Bank's money desk to "provide more ample funds and encourage the uncollateralised overnight call rate to move as low as possible." No number was given, because no number was the point. Within weeks the rate was trading at two or three hundredths of one per cent, which is to say zero.

Promising to Be Irresponsible

A zero policy rate exhausts the conventional instrument, and the debate that followed produced most of the vocabulary that central banks would use a decade later. Paul Krugman had argued the previous year that Japan was in a genuine liquidity trap, and that the way out was not more money today but a credible promise of inflation tomorrow: the central bank, he wrote, had to "credibly promise to be irresponsible" (Krugman, 1998). Ben Bernanke, then a Princeton economist, published a paper in 2000 titled "Japanese Monetary Policy: A Case of Self-Induced Paralysis?" and answered his own question in the affirmative.

Hayami took a version of the advice. In April 1999 he told a press conference that the Bank would hold the overnight rate at zero "until deflationary concerns are dispelled." That sentence was the first modern example of what is now called forward guidance, and its mechanism is subtle: a promise about the future path of overnight money pulls down yields at one, two and five years today, because a five-year rate is an average of expected overnight rates. Eggertsson and Woodford later formalised the point, showing that at the zero bound the only remaining lever is a commitment about the future, and that the commitment works precisely to the extent that it is understood to bind the central bank against its own later preferences (Eggertsson and Woodford, 2003).

Hayami had no appetite for being bound. He believed persistently that zero rates damaged the functioning of money markets, weakened the discipline that failing borrowers ought to face, and punished Japanese savers, and he said so repeatedly while the policy he disliked remained in force.

Eighteen Months, Then a Mistake

On 11 August 2000 the Policy Board lifted the zero interest rate policy and returned the call rate target to 0.25 per cent. Core consumer prices were still falling. The government, which under the 1998 law retains the right to attend meetings and request a postponement of a vote, formally asked the Board to delay. The request was rejected, and the increase passed seven votes to two. It was the first and only time a Japanese government has used that power, and the Bank's decision to override it was widely read in Tokyo as the new institution proving its independence at the first available opportunity.

Timing was against them. Japan's own business-cycle peak fell in October 2000, and the collapse of the American technology market — traced in the dot-com bubble of 1995–2000 — removed the export demand the recovery had been resting on. By February 2001 the Board was cutting again and had introduced a standing lending facility to cap money-market rates. The August increase had lasted seven months, and it cost the Bank something harder to rebuild than a quarter point: every subsequent promise to keep rates at zero now had to be weighed against the memory that this Board had broken one.

19 March 2001

Faced with a rate it could not lower further and a commitment markets had stopped believing, the Policy Board changed the variable it was steering. On 19 March 2001 it announced that the operating target of money market operations would no longer be the overnight call rate but the outstanding balance of current accounts held by financial institutions at the Bank of Japan, to be maintained at around ¥5 trillion against required reserves of roughly ¥4 trillion.

Three features of that decision made it the template for everything that followed. First, the target was a quantity, not a price: the Bank committed to flooding the banking system with reserves far beyond anything the payment system needed. Second, the commitment was written against an observable statistic rather than a feeling — the policy would run until the year-on-year change in core consumer prices was "zero per cent or above on a sustainable basis," which removed the discretion that had destroyed the previous promise. Third, to supply reserves on that scale the Bank undertook to increase its outright purchases of long-term Japanese government bonds, then running at ¥400 billion a month.

The Bank called it ryōteki kinyū kanwa. The English rendering, quantitative easing, entered the language on that day.

Escalation was steady and, at each step, insufficient. The target went to ¥6 trillion in August 2001, was raised again in the week after the 11 September attacks, and reached a range of ¥10–15 trillion that December. Monthly bond purchases climbed in parallel, from ¥400 billion to ¥1.2 trillion by October 2002 — an annual run rate above ¥14 trillion in a government bond market the Bank was now the dominant buyer in.

Fukui, Resona, and the VaR Shock

Toshihiko Fukui replaced Hayami on 20 March 2003 and moved faster in his first ten months than his predecessor had in two years, raising the reserve target four times. His arrival coincided with the resolution of the banking problem that had sat underneath the whole deflation. Resona Bank, the country's fifth-largest, was found in May 2003 to have a capital ratio around half the required level once its deferred tax assets were properly examined; the government injected ¥1.96 trillion of public money on 17 May without wiping out shareholders. Japanese bank shares rallied on the news, which told the authorities something useful — the market had concluded that no large Japanese bank would now be allowed to fail on its own.

Then the bond market broke in the opposite direction. Ten-year Japanese government bond yields had fallen to 0.430 per cent on 12 June 2003, a record low, as banks piled into duration with reserves they had no borrowers for. Over the following three months the yield rose to roughly 1.6 per cent. Japanese banks ran their bond books against value-at-risk limits calibrated on the preceding period of calm, so the initial sell-off pushed risk measures through their ceilings, which forced sales, which pushed yields higher and risk measures higher still. Tokyo named it the VaR shock, and it is a close cousin of the mechanism that later broke the long end of the British curve in the 2022 gilt crisis.

Watching a quarter of its yield-curve effect vanish in a summer, the Bank rewrote its commitment. On 10 October 2003 it published explicit exit conditions: core consumer prices must have registered zero or above for several months, Policy Board members must forecast that they would not fall back below zero, and even then the Bank might continue if circumstances warranted. Vagueness had been replaced with a public test the Bank could be held to.

Currency policy was doing separate work. Between January 2003 and March 2004 the Ministry of Finance sold roughly ¥35 trillion of yen for dollars, the largest intervention campaign any government has ever run, much of it unsterilised into a banking system already awash with reserves.

What It Actually Achieved

Reserves at the Bank rose roughly sevenfold between March 2001 and January 2004. Bank lending to the private sector did not rise at all; it contracted in every one of those years. Broad money grew at around 2 per cent a year throughout, which is to say that the textbook money multiplier had stopped working, because reserves earned no interest but also faced no shortage of willing holders among banks with impaired balance sheets and no creditworthy borrowers.

Hiroshi Ugai's survey of the empirical literature, published by the Bank itself in 2007, reached a conclusion that has held up. The clearly demonstrable effect of quantitative easing was on the yield curve, and it came from the commitment rather than the quantity: by making the zero rate credible for longer, the policy flattened yields out to several years and lowered borrowing costs across the economy (Ugai, 2007). Effects running through portfolio rebalancing — the channel by which stuffing banks with reserves was supposed to push them into riskier assets — were weak and hard to identify. Oda and Ueda reached a similar verdict using a term-structure model, attributing most of the decline in medium-term yields to expectations about the policy rate rather than to the size of the Bank's balance sheet (Oda and Ueda, 2007).

Quantitative easing also did something no model captures well. Interbank markets in Japan functioned normally throughout a period in which several major banks were arguably insolvent, because no institution could plausibly claim to be short of yen. The Japan premium disappeared. Whether or not the policy raised prices, it made a banking crisis quiet.

Exit, and Export

Core consumer prices turned positive in the autumn of 2005 and stayed there. On 9 March 2006 the Board declared the conditions met and terminated quantitative easing, returning to the overnight call rate as its operating target and draining more than ¥20 trillion of excess reserves over the following months without disturbance. The rate itself went to 0.25 per cent on 14 July 2006, ending seven years of zero, and to 0.5 per cent in February 2007. It would be seventeen years before Japanese policy rates saw the far side of one per cent.

Foreign central banks had spent the intervening years insisting the Japanese experience was a Japanese problem — a story about bad loans, zombie firms and a supervisory culture that postponed every reckoning. That reading did not survive September 2008, described in the 2008 financial crisis, after which the Federal Reserve, the Bank of England and eventually the European Central Bank adopted the zero floor, the balance-sheet expansion and the conditional commitment in roughly the order Tokyo had discovered them.

Bernanke, by then chairman of the Federal Reserve, went to the London School of Economics in January 2009 to explain what his institution was doing, and spent part of the speech insisting that the Fed's programme was credit easing rather than quantitative easing, since it targeted the composition of assets held rather than the quantity of reserves created. The distinction was real, and nobody kept it. Within a year the American policy was universally called QE, in the English phrase invented to translate a Japanese one.