10.5 Japan's Lost Decades
Japan’s stock and land bubble burst in 1990 and 1991, leaving firms whose debts were bigger than the value of their assets. Instead of borrowing, they used their cash flow to pay down debt, so near-zero interest rates failed to revive lending. The economist Richard Koo calls this a balance sheet recession. The Nikkei took 34 years to regain its December 1989 record.
Why it matters: When everyone repays debt at once, cheap money alone cannot restart an economy.
Summary: Japan’s stock and land bubble burst in 1990–91, leaving firms whose debts exceeded the value of their assets. They used their cash flow to repay debt instead of borrowing, so near-zero rates failed to revive credit: Richard Koo’s balance sheet recession. The Nikkei took 34 years to regain its December 1989 record.
- A firm with ¥100 billion of property and ¥70 billion of debt has −¥30 billion of equity after a 60% fall and needs 8.4 years of cash flow to restore a 30% equity ratio.
- Koo put Japan’s commercial real estate fall at 87% and its corporate net debt repayment at up to ¥30 trillion a year.
- The Nikkei fell 81.9% from 38,915.87 (December 29, 1989) to its March 2009 low.
- It first closed above the 1989 record on February 22, 2024.
- The Bank of Japan’s policy rate target was around 1.25% from September 2026.
Through the 1980s, Japanese real estate and stock prices rose to extraordinary levels — at the peak, the land underneath the Imperial Palace in Tokyo was reportedly valued at more than the entire state of California. When this asset bubble burst in 1990–91, Japan entered a multi-decade period of anemic growth, persistent mild deflation, and near-zero interest rates that has come to define an entire economic archetype: the balance sheet recession, where over-indebted companies and households prioritize paying down debt over new spending or investment for years, blunting the effectiveness of conventional monetary policy (Volume I’s Part 2) even at very low rates.
Japan’s experience became the direct reference case invoked repeatedly after 2008 and again during the pandemic, as central banks worldwide worried about their own economies falling into a similar low-growth, low-inflation, debt-overhang trap — and Japan’s own eventual, tentative return to modest inflation and rate normalization in the 2020s is watched globally as a live test of how, and whether, such an entrenched pattern can be escaped.
Volume I’s Part 9 tells the full story of the bubble, the zombie lending and the deflation that followed; this box isolates the balance-sheet arithmetic behind the term used by the economist Richard Koo. Take an illustrative Japanese company in 1990 with ¥100 billion of property financed by ¥70 billion of debt, so equity is ¥30 billion. Property then falls 60%.
| Step | Formula | Result |
|---|---|---|
| Assets after the fall | 100 × (1 − 0.60) | ¥40 billion |
| Equity after the fall | 40 − 70 | −¥30 billion (technically insolvent) |
| Debt consistent with the old 30% equity ratio | 40 × (1 − 0.30) | ¥28 billion |
| Debt to repay | 70 − 28 | ¥42 billion |
| Years of repayment at ¥5 billion of free cash flow a year | 42 ÷ 5 | 8.4 years |
The business still earns money, so it is not bankrupt in cash terms, but every yen of cash flow goes to paying down debt for 8.4 years. A lower interest rate shortens that slightly; it does not make the firm want to borrow. When thousands of firms do this at once, private demand for credit disappears even at zero rates. Koo, testifying to the US House Financial Services Committee on July 22, 2010, put Japan’s commercial real estate fall at 87% nationwide and noted that its corporate sector was a net repayer of debt for about ten years after 1995, at up to ¥30 trillion a year (about 6% of GDP); government borrowing filled the gap. Prices show the scale: the Nikkei 225 closed at a record 38,915.87 on December 29, 1989, fell 1 − 7,054.98 ÷ 38,915.87 = 81.9% to its March 2009 low, and first closed above the 1989 record on February 22, 2024, at 39,098.68, a price gain of 0.5% over 34 years (dividends excluded).
Reading rule: after an asset bust, judge whether rate cuts will work by the private sector’s balance sheets, not by the level of rates. If a large share of firms or households have liabilities above the market value of their assets, expect them to repay debt whatever the rate, and expect recovery to need either fiscal spending, fast recognition of bank losses, or both. If balance sheets are sound and the problem is only confidence, conventional rate cuts usually work. Do not apply the Japan template to a country whose bust was small relative to private debt.
Buying the “cheap” market after the first big fall. A buyer at half the Nikkei’s 1989 peak, 0.5 × 38,915.87 = 19,458, still faced a further fall of 1 − 7,054.98 ÷ 19,458 = 63.7% to the 2009 low, and a buyer at the peak waited 34 years to see the index level again, a compound price return of roughly 0% a year. A 50% fall says nothing about value when earnings and land prices are themselves deflating. Value the market on current earnings and the debt overhang, not on its distance from the peak.
What is a balance sheet recession?
It is a downturn in which companies and households, whose debts exceed the value of their assets after a crash, use their income to repay debt instead of spending or investing. Because they are minimizing debt rather than maximizing profit, low interest rates do little to revive borrowing. Richard Koo used the term to describe Japan after 1990.
How long did it take the Nikkei to recover?
Thirty-four years. The index closed at 38,915.87 on December 29, 1989, and did not close higher until February 22, 2024, when it reached 39,098.68. At its March 2009 low of 7,054.98 it was down 81.9% from the peak.
Has Japan escaped its low-rate era?
Partly. The Bank of Japan moved its policy rate from −0.1% to a 0 to 0.1% range in March 2024, ending eight years of negative rates, and has raised it in steps since; on September 18, 2026, it set the target for the overnight call rate at around 1.25%. Whether inflation and wage growth hold up is still being tested.
After Japan’s bubble burst, firms with debts above their asset values repaid debt instead of borrowing, so near-zero rates could not revive credit: the balance sheet recession. The Nikkei fell 81.9% to its 2009 low and took 34 years to regain its 1989 record.
Six questions on this chapter. Decide on your answer first, then click “Reveal Answer.”
1. A firm has ¥100 billion of property and ¥70 billion of debt. Property falls 60%. What is its equity?
- ¥30 billion
- −¥30 billion
- ¥12 billion
- −¥42 billion
Reveal Answer
Answer: B. Assets become 100 × 0.4 = ¥40 billion, and 40 − 70 = −¥30 billion.
2. Why did near-zero interest rates fail to restart borrowing in Japan’s balance sheet recession?
- Banks were barred from lending below 1%
- The central bank kept long rates above 5%
- Firms preferred issuing shares at high prices
- Firms used cash flow to repay debt
Reveal Answer
Answer: D. Firms with liabilities above their asset values minimize debt, so cheaper credit does not tempt them.
3. Roughly how long did it take the Nikkei 225 to close above its December 29, 1989, record?
- 19 years
- 12 years
- 34 years
- 48 years
Reveal Answer
Answer: C. It first closed higher on February 22, 2024, at 39,098.68.
4. An investor bought the Nikkei at half its 1989 peak, about 19,458. What further fall did that investor face to the 2009 low of 7,054.98?
- About 64%
- About 36%
- About 50%
- About 82%
Reveal Answer
Answer: A. 1 − 7,054.98 ÷ 19,458 = 63.7%; 81.9% is the fall from the peak itself.
5. Worked problem: A firm has ¥100bn of property and ¥70bn of debt. Property falls 50%. What is its equity?
Reveal Answer
Answer: Property = ¥50bn, debt = ¥70bn, so equity = −¥20bn.
6. Worked problem: If the firm generates ¥4bn of cash flow a year, how long does it take to repay that gap?
Reveal Answer
Answer: ¥20bn ÷ ¥4bn = 5 years, during which it repays debt instead of investing or borrowing.
10.6 The Dot-Com Crash (2000)
The Nasdaq Composite fell 77.9%, from 5,048.62 on March 10, 2000, to 1,114.11 on October 9, 2002, and did not regain its peak until April 2015. Prices had required earnings growth that few firms could deliver. The conflicts of interest behind the hype led to a $1.4 billion research settlement in 2003.
Why it matters: A price can run far ahead of what a business can ever earn.
Summary: The Nasdaq Composite fell 77.9%, from 5,048.62 on March 10, 2000, to 1,114.11 on October 9, 2002, and did not regain its peak until April 2015. Prices had required earnings growth few firms could deliver, and the conflicts behind the hype led to a $1.4 billion research settlement in 2003.
- At 100 times earnings, a stock needs 17.5% annual earnings growth for ten years just to hold its price as the multiple falls to 20.
- With a 10% required return, the needed growth is 29.2% a year for a decade.
- Ten firms paid $1,387.5 million in April 2003 and separated research from investment banking.
- Pástor and Veronesi argue high uncertainty can justify high prices; Ofek and Richardson link the collapse to short-sale limits and lockup expiries.
- A buyer at the 2002 low needed a 353% gain to break even from the peak.
Through the late 1990s, internet-related companies — many with minimal revenue and no path to profitability clearly articulated — reached enormous valuations purely on narrative and growth expectation, echoing the South Sea Bubble’s genuine-business-versus-price detachment nearly three centuries later. Comparable company multiples (Part 1: Corporate Finance & Valuation) became almost meaningless across the sector, as analysts substituted metrics like “eyeballs” and “page views” for the profit and cash flow measures Volume I’s Part 5 identified as the honest measures of business quality. When the Nasdaq index peaked in March 2000 and began an extended decline, it ultimately fell 77.9% from peak to trough by October 2002, and a large share of the era’s most hyped companies disappeared entirely.
The dot-com crash is the South Sea Bubble with a technology narrative instead of a South American trade monopoly — the specific story changes each cycle, but the mechanism is identical: an important underlying technological or economic shift (the internet was, in fact, transformative) gets wildly overpriced in the short run, because investors correctly identify the destination but badly misjudge both the timeline and which specific companies would actually survive to reach it.
Run the valuation backward (Part 1: Corporate Finance & Valuation). Suppose a stock trades at 100 times earnings and a mature company of its kind deserves 20 times. For the price merely to stay flat while the multiple falls to 20 over ten years, earnings must grow (100 ÷ 20)1/10 − 1 = 17.5% a year for a decade. If you also want a 10% annual return, they must grow (1.1010 × 100 ÷ 20)1/10 − 1 = 29.2% a year for a decade. Few companies have ever done that, and a sector full of firms priced this way assumes nearly all of them will. That is the arithmetic of 2000: the Nasdaq Composite closed at 5,048.62 on March 10, 2000, fell to 1,114.11 on October 9, 2002, a loss of 1 − 1,114.11 ÷ 5,048.62 = 77.9%, and did not close above its peak until April 23, 2015.
The conflicts that fed the narrative were then priced. On April 28, 2003, the SEC, the New York Attorney General, NASD, NYSE and state regulators settled with ten investment firms whose research analysts had promoted the stocks their bankers were selling: $487.5 million of penalties, $387.5 million of disgorgement, $432.5 million for independent research and $80 million for investor education, $1,387.5 million in all. The firms had to separate research from investment banking and stop paying analysts on banking revenue.
Rational but uncertain: Ľuboš Pástor and Pietro Veronesi (NBER working paper 10581, 2004) answered their title question “Was there a Nasdaq bubble in the late 1990s?” with “not necessarily”: a firm’s value rises with uncertainty about its future profitability, and the uncertainty needed to match peak valuations was “high but plausible.” Mispricing that could not be corrected: Eli Ofek and Matthew Richardson (Journal of Finance, 2003) found internet stocks were hard to sell short, so the most bullish buyers set prices, and linked the collapse to the expiry of lockup agreements that had kept insiders from selling. Both views agree on the trigger mechanism: when supply of shares or doubt about profits rose, prices had no floor.
For any stock or sector priced on a story, compute the earnings growth the price requires: (current multiple ÷ mature multiple × (1 + required return)years)1/years − 1. If the answer is above about 20% a year for ten years, treat the position as a bet on one of a few winners and size it as you would an option, small enough to lose entirely. If several firms in one sector all need that growth, assume most will fail. The test is weaker for firms already earning high returns on capital with long reinvestment runways.
Being right about the technology and wrong about the price. The internet did transform commerce, yet a buyer of the Nasdaq Composite at its March 10, 2000, close waited fifteen years, to April 23, 2015, to see that level again, and at the October 2002 low needed a 5,048.62 ÷ 1,114.11 − 1 = 353% gain just to break even. Pay for the earnings a company can plausibly reach, not for the size of the trend it belongs to.
How much did the Nasdaq fall in the dot-com crash?
The Nasdaq Composite fell 77.9%, from a close of 5,048.62 on March 10, 2000, to 1,114.11 on October 9, 2002. It first closed above its 2000 peak on April 23, 2015.
What caused the dot-com bubble to burst?
No single event. Valuations required growth few firms could deliver, and the supply of shares rose as insiders’ lockups expired; Ofek and Richardson link the collapse to that insider selling. Once profits failed to appear, funding for loss-making firms dried up and many failed.
Did regulators respond to the dot-com crash?
Yes. The April 2003 global research settlement made ten firms pay about $1.4 billion and separate research from investment banking. The accounting scandals that surfaced in the same downturn led to the Sarbanes-Oxley Act of 2002, whose CFO certification rule Section 7.6: CFO and Other C-Suite Hiring — Who Actually Decides explains.
Dot-com valuations required earnings growth of 17.5% to 29% a year for a decade, and the Nasdaq fell 77.9% between March 2000 and October 2002, regaining its peak only in 2015. Researchers still debate whether the prices were irrational or just very uncertain, but the 2003 research settlement priced the analyst conflicts that fed them.
Six questions on this chapter. Decide on your answer first, then click “Reveal Answer.”
1. A stock trades at 100 times earnings. If the multiple falls to 20 over ten years and the price stays flat, what annual earnings growth is required?
- About 8.0%
- About 29.2%
- About 17.5%
- About 50.0%
Reveal Answer
Answer: C. (100 ÷ 20)^(1/10) − 1 = 17.5%; adding a 10% required return raises it to 29.2%.
2. How far did the Nasdaq Composite fall from March 2000 to October 2002?
- About 89%
- About 78%
- About 49%
- About 64%
Reveal Answer
Answer: B. 1 − 1,114.11 ÷ 5,048.62 = 77.9%.
3. What did the April 2003 research analyst settlement require?
- Separating research from banking
- A ban on IPOs by unprofitable firms
- Prison terms for analysts with buy ratings
- A cap on P/E ratios for new listings
Reveal Answer
Answer: A. The firms paid about $1.4 billion and had to wall off research and stop paying analysts on banking revenue.
4. What did Ofek and Richardson link to the collapse of internet stock prices?
- The end of the Global Settlement in 2003
- Rising margin requirements set by the Fed
- A ban on short selling of technology stocks
- Expiring lockups that let insiders sell
Reveal Answer
Answer: D. They found short-sale constraints let optimists set prices and tied the fall to insider selling as lockups expired.
5. Worked problem: A stock trades at 80× earnings. What annual earnings growth for ten years is needed to hold its price as the multiple falls to 20×?
Reveal Answer
Answer: Growth = (80 ÷ 20)1/10 − 1 = 14.9% a year.
6. Worked problem: At 60× falling to 20× over ten years, with a 10% required return, what growth is needed?
Reveal Answer
Answer: Growth = (60 ÷ 20)1/10 × 1.10 − 1 = 22.8% a year.
- Koo, testimony to the House Financial Services Committee, July 22, 2010 — Balance sheet recession figures
- SEC press release 2003-54, research analyst settlement — $1.4 billion settlement
- FRED, NASDAQ Composite Index (NASDAQCOM)