The Dot-Com Bubble: A 78 Percent Fall and a 15-Year Recovery

Key Takeaways

• The Nasdaq Composite fell 77.9 percent from its 10 March 2000 close of 5,048.62 to its 9 October 2002 close of 1,114.11, over 647 trading sessions.
• Recovery took 5,522 calendar days. The index first closed back above 5,048.62 on 23 April 2015.
• The broad market fell far less. The S&P 500 declined 49.1 percent and recovered on 30 May 2007, about seven years after its own peak. The Dow Jones Industrial Average fell 37.8 percent and recovered in October 2006.
• The spread between those recoveries, fifteen years against seven, is the single most important number on this page. It is a measure of concentration, not of technology.
• The Federal Reserve had been tightening into the peak: its target rate rose from 5.00 percent in June 1999 to 6.50 percent in May 2000, then fell to 1.75 percent through eleven cuts during 2001.
• The most-cited warning signal fired after the top. The 10-year Treasury yield first closed below the 3-month yield in 2000 on 7 April, nearly a month after the Nasdaq Composite peak.

What Happened in the Dot-Com Bubble?

Between the mid-1990s and March 2000, capital moved into companies whose business models depended on widespread internet adoption. The underlying thesis was correct. Internet adoption did become widespread, and it did create enormous economic value. What the market got wrong was not the technology forecast. It was how much of that value would accrue to the specific listed companies then trading, and at what price it made sense to buy that claim.

The unwind was not a crash in the 1987 sense. It was a long, uneven decline across two and a half years, punctuated by rallies that each looked like a bottom.

Chronology

Dated events with index closes. Index closes computed from daily closing values; policy rates from Federal Reserve records.

Two features of that timeline deserve attention. The recession ended in November 2001 and the market fell for another eleven months, bottoming in October 2002. And the three indexes peaked ten weeks apart and recovered nearly nine years apart. Both facts are inconvenient for simple narratives about how markets and economies relate.

What Did the Setup Look Like Before the Peak?

A genuinely correct technology thesis. This is what makes the dot-com bubble instructive rather than merely cautionary. The internet did transform commerce, communication and media. Anyone who forecast that in 1999 was right. Being right about it was insufficient, and that is the point.

Extreme concentration of the advance. The Nasdaq Composite peaked at 5,048.62 and the Dow at 11,722.98 within ten weeks of each other, and the Nasdaq subsequently fell more than twice as far. That divergence is a direct measurement of how concentrated the advance had been.

A tightening cycle already underway. The Federal Reserve's archived record shows six increases from 30 June 1999 to 16 May 2000, taking the target rate from 4.75 percent to 6.50 percent. The market rose through the first five of them. As in 1929, official tightening aimed at an overheating market did not stop the advance on any timetable an investor could use.

A strong labor market. Unemployment reached 3.8 percent in April 2000, its low for that cycle, one month after the Nasdaq peak. It then rose steadily to 6.3 percent by June 2003. As with 2022, the labor market gave no advance warning and was still improving at the top.

Very high short rates. On 10 March 2000 the 3-month Treasury bill yielded 5.87 percent and the 10-year note 6.39 percent. That is an important and under-discussed piece of context: cash paid nearly 6 percent, so the opportunity cost of not owning equities was real. Investors were not reaching for return because there was no alternative.

How Far Did Prices Fall?

Peak-to-trough decline by index. Computed from daily closing values; close to close, price only.

Look at the spread. Three United States equity indexes, over the same window, with declines of 77.9 percent, 49.1 percent and 37.8 percent. The difference between holding the technology-weighted index and the industrial average was 40 percentage points of drawdown, from the same starting date to the same ending date.

The calendar-year figures show how prolonged it was. Computed from first to last close of each year, the S&P 500 fell 9.3 percent in 2000, 10.5 percent in 2001 and 23.8 percent in 2002. Three consecutive down years is unusual and it is what distinguishes this from a single sharp decline: there was no single moment of capitulation to react to, only a grinding sequence in which each year's loss was survivable and the cumulative loss was not.

The recovery arithmetic

A 77.9 percent decline requires the surviving capital to multiply by about 4.5 times to break even. That is why the Nasdaq Composite took fifteen years while the S&P 500, needing to roughly double from its low, took seven. The relationship between depth and recovery time is not linear, and it becomes punishing quickly. The same arithmetic is set out in more detail in the 1929 case study.

Which Warning Signs Were Visible in Advance, and Which Only in Hindsight?

Signals classified by whether they were usable at the time.

The honest reading is that the valuation observation was correct and useless for timing, and the concentration observation was correct and useful for sizing. Those are different kinds of information and they support different actions. Noting that prices were high justified no particular trade. Noting that a portfolio's outcome depended almost entirely on one sector justified a concrete change in position size, available at any point, requiring no forecast at all.

Hindsight check. The retrospective consensus is that the bubble was obvious. What that account leaves out is that the same people making the argument in 1998 watched the Nasdaq Composite rise about 88 percent from mid-1999 to the peak alone. Being early is indistinguishable from being wrong in the account statement, and it is what the experience actually felt like. It also leaves out the harder question, which was never "are prices high" but "which of these businesses will still exist," and the answer to that was genuinely not available. Performance chasing covers the pressure that operated on anyone positioned defensively through 1999.

Why Did the Recovery Take So Long?

Three mechanisms, and the third is the one that generalizes.

The depth made the arithmetic brutal. Recovering from a 77.9 percent decline requires a 355 percent gain. Even a strong subsequent decade does not necessarily deliver that, and in this case it took two full market cycles.

The index composition changed underneath. Many of the companies that constituted the March 2000 peak did not participate in the recovery, because they no longer existed as independent listed businesses. An index level returning to 5,048.62 in 2015 was not the same collection of companies returning to their old prices. It was a substantially different index reaching the same number.

Concentration converted a sector problem into a portfolio problem. This is the durable lesson. An investor holding the Dow was made whole in six and a half years. An investor holding the Nasdaq Composite waited fifteen. Neither made a market-timing decision. The difference came entirely from a composition choice made before anything happened, and it was the largest single determinant of the outcome.

It is worth stating the uncomfortable version explicitly. The technology thesis was right, and the technology-weighted index still took fifteen years to break even. Being correct about the direction of the world does not protect an investor from the price paid to express that view, and it does not protect against the specific companies chosen to express it. That is a general property of concentrated positions, not a property of the internet.

How Did the Federal Reserve Respond?

The Federal Reserve's archived record shows a tightening cycle running straight into the peak, followed by the most aggressive single year of easing in its modern record to that point.

Federal funds target rate, June 1999 to November 2002, from the Federal Reserve Board's archived record of open market operations.

Eleven cuts in 2001, taking the target from 6.50 percent to 1.75 percent, and the market fell throughout and for another ten months afterwards. This is the clearest counterexample in this library to the belief that rate cuts support equity prices in any reliable near-term way. The easing was large, fast and continuous, and the Nasdaq Composite lost roughly half its remaining value while it was happening.

The recession itself was mild by the standards of this library. The NBER dates it from March 2001 to November 2001, eight months, and unemployment peaked at 6.3 percent in June 2003, well below the 10 percent of 2009 or the 14.8 percent of 2020. A mild recession accompanied the deepest index drawdown here after 1929, which is another reminder that equity drawdowns and economic contractions are related but not proportional. Federal Reserve policy rates and forward guidance covers why the transmission is slower and less direct than the popular account assumes.

How Long Did the Recovery Take?

Time from peak close back to that same close. Computed from daily closing values; price only, dividends excluded.

There is a detail in that table that is easy to miss and worth dwelling on. The Dow and the S&P 500 recovered in 2006 and 2007, which is months before both began the 2007 to 2009 decline covered in the 2008 case study. An investor in the broad market spent seven years getting back to even and then immediately gave back more than half of it. From the March 2000 peak, the S&P 500 did not durably hold above that level until well into the following decade.

That sequence is the strongest available argument against measuring an investment experience by a single peak-to-recovery statistic. The recovery date says the drawdown ended in 2007. The lived experience of the period from 2000 to 2013 was of a market that twice returned to the same level and twice failed to hold it.

As elsewhere in this library, these are price measures. Dividends reduce the time to break even, and this page does not state a total-return recovery date because none was verified from a source in this session. Inflation extends it in real terms.

What Was Specifically Different About the Dot-Com Bubble?

The underlying thesis was correct. Unlike a decline caused by a credit failure or an external shock, the story investors believed in 1999 largely came true. The failure was in price and in company selection, not in the forecast. That combination is unusual and it is what makes this episode uniquely useful for thinking about concentrated positions.

The dispersion across indexes was extreme. A 40 percentage point spread in drawdown between the technology-weighted index and the industrial average, and a nine-year spread in recovery, over identical dates. No other episode in this library comes close.

The banking system was unaffected. No lender of last resort function was needed, no institutions failed in a systemically important way, and the recession was mild. This was an equity valuation event with limited transmission into credit.

Aggressive easing did not stop the decline. Eleven cuts in 2001 and the market kept falling into October 2002. This is the counterexample to the 2020 pattern.

Cash paid nearly 6 percent. A 3-month Treasury bill yielding 5.87 percent at the peak meant a genuine, low-risk alternative existed. Every subsequent episode in this library occurred with materially lower short rates until 2022.

Common Myths About the Dot-Com Bubble

"The internet was overhyped." It was not. The technology delivered more than most 1999 forecasts predicted. What failed was the price paid and the selection of which companies would capture the value. Treating this as a cautionary tale about technology optimism misidentifies the error entirely.

"Everyone knew it was a bubble." The valuation argument was widely made from 1998. Computed from daily closes, the Nasdaq Composite rose more than threefold from the start of 1998 to the March 2000 peak, most of it after that argument had become commonplace. Being right about the eventual outcome while being two years early is a very different experience from being right, and most of the people remembered as prescient are remembered selectively.

"The recession caused the crash." The Nasdaq Composite peaked in March 2000 and the NBER dates the recession from March 2001, a full year later. The recession ended in November 2001 and the market fell for eleven more months. The two events overlap without one explaining the other.

"Rate cuts stopped the decline." Eleven cuts during 2001 took the target from 6.50 percent to 1.75 percent, and the Nasdaq Composite fell throughout and into October 2002. If anything this episode is evidence that easing into a valuation unwind does relatively little in the near term.

"It recovered eventually, so buying the top was fine." Fifteen years to break even on price, from a peak that a working-life investor might have reached in their forties, is not a minor inconvenience. Whether it was fine depends entirely on the holder's horizon and whether they were adding or withdrawing during those fifteen years.

What a Reader Can Actually Carry Forward

The dot-com bubble is the most useful episode in this library for anyone holding a concentrated position, and concentration is far more common than it appears. A market-capitalization-weighted index concentrates automatically as its largest holdings appreciate, without the holder making any decision at all.

What generalizes

• Being right about a trend does not make you right about a price. The internet thesis was correct and the Nasdaq Composite still took fifteen years to break even. Any investment case that rests on a correct forecast about the world needs a separate answer to the question of what price makes that forecast worth owning.
• Concentration determines dispersion of outcomes. Same country, same period, same trough date, 40 percentage points of difference in drawdown. That gap was decided by a composition choice made in advance, and it dominated everything else.
• Index composition changes underneath the level. An index recovering its old number is not the same companies recovering their old prices. Long-horizon comparisons of index levels quietly include survivorship.
• Rate cuts do not reliably stop a valuation unwind. Eleven cuts in one year, and the decline continued for another ten months afterwards.
• Recovery dates hide the lived path. The S&P 500 recovered in May 2007 and then fell 56.8 percent. A single peak-to-recovery statistic can describe an experience that was nothing like what the statistic implies.

What does not generalize

• The 77.9 percent decline. It required an extreme concentration into a single sector at an extreme price. It is not a base rate for anything.
• The fifteen-year recovery. One observation, driven by the depth and by composition change.
• The mild recession alongside a severe drawdown. That relationship varies enormously and 2008 inverted it.
• Cash paying nearly 6 percent. The opportunity cost calculation facing an investor in 2000 was structurally different from the one facing an investor in 2020.

The one question worth asking now

The practical exercise is to look up how much of your equity exposure sits in the largest handful of holdings, including inside every fund you own, and then ask what happens to your plan if that specific group underperforms for a decade. This requires no forecast about which companies will win, which is fortunate, because that is the question nobody answered correctly in 1999 either. Concentration is measurable today. The outcome it produces is not, and that asymmetry is the entire content of this episode. Stress testing and scenario analysis is the structured version of that exercise.

References

Every figure on this page was verified against the following sources, each retrieved on 23 August 2026:

• National Bureau of Economic Research: US Business Cycle Expansions and Contractions: the March 2001 peak and November 2001 trough dating the recession.
• Federal Reserve Board: Open Market Operations Archive: every federal funds target change and resulting level in the 1999 to 2002 table above, including the six increases from 30 June 1999 to 16 May 2000 and the eleven reductions during 2001.
• US Department of the Treasury: Daily Treasury Par Yield Curve Rates: the 3-month and 10-year yields quoted for 1999 and 2000, the absence of any 3-month to 10-year inversion during 1999, and the 7 April 2000 first inverted close of that year, computed from the published daily series.
• US Bureau of Labor Statistics: Unemployment Rate, Series LNS14000000: the 3.8 percent April 2000 cycle low and the 6.3 percent June 2003 peak.

Figures deliberately not stated. This page gives no count of internet company failures, no aggregate figure for capital raised or lost in the period, no price-to-earnings ratio for any index at the peak, and no total-return recovery date, because no source verified in this session supplied them. Where the shape of an event is known but the number is not, the shape is described and the number is left out rather than estimated.

Method note: index peak, trough, decline and recovery figures labeled as computed were derived by Swoopr Investment from daily closing values of the named index, retrieved from the Yahoo Finance historical chart API on 23 August 2026. Drawdowns are measured close to close, not intraday, so the intraday low of any episode is lower than the trough shown. Recovery means the first daily close at or above the prior peak close, price only, with no dividends reinvested. Figures labeled total return are computed instead from dividend-adjusted closing prices and are stated as such wherever they appear.

This is educational content about a historical episode. It is not investment advice, it is not a forecast, and nothing here should be read as a claim about how any future market decline will behave.

Frequently Asked Questions