Does Tech Fall Harder in Bear Markets? US Tech vs the Market, 1971–2026

The Tech/Market Ratio Against Real Interest Rates, 1971–2026

How to Read the Ratio

The ratio tracks the total-return index of the US technology sector (Kenneth R. French’s HiTec portfolio) divided by that of the whole US market, both at month-end and normalised so February 1971 = 100. By May 2026 the ratio reaches 129, indicating tech has cumulatively returned 1.29× as much as the market since the start of the series. Over the period it moves between about 59 (September 2002 and July 2006) and 149 (February 2000), a range of 2.5×.

The real rate used is the ex-post real 10-year Treasury yield, computed as the monthly average DGS10 minus contemporaneous CPI year-over-year. This is a 55-year-consistent measure; TIPS-based real yields are only available from 2003. See our US Real Interest Rates Dataset (1962–Present) for the underlying real-rate series.

Federal Reserve tightening cycles are defined here as periods where the Fed Funds rate rose by ≥150 basis points without an intervening cut of more than 50 basis points. Eight such cycles are identified between 1971 and 2024. How the ratio relates to the real yield before and after the mid-1990s is examined in the real-rate section below.

The Duration Framing: and Where It Comes from

The dominant narrative for the Nasdaq’s behaviour during periods of monetary tightening holds that technology equities are long-duration assets. A high share of expected cash flows accrues in the distant future, so any increase in the discount rate compresses present value more than for an index dominated by mature, cash-generating businesses. For the mechanics, see our primer on bond duration. By this framing, the Nasdaq would be expected to fall harder than the S&P 500 whenever real interest rates rise, and benefit disproportionately when they fall.

Measured with the technology sector rather than the Nasdaq Composite, the data support this framing only in part. Tech’s cumulative total return, +47,238% versus +36,464% for the market from February 1971 to May 2026, came entirely after 1995 (the ratio fell before 1995) and not in negative real-rate regimes: months that began with a negative ex-post real yield took the ratio down (×0.90 over 124 months). And in the six tech bear markets ≥20%, tech’s drawdown exceeded the market’s in 3 of 6 cases, with an average ratio across all six of 1.22×. The “falls harder” pattern is real but conditional: it holds in 1990, 2000–02 and 2021–22, not in 1973–74, 1987 or 2007–09.

Six Tech Bear Markets, 1971–2026

A tech bear market is defined here as any peak-to-trough decline of 20% or more in the month-end total-return index of the US technology sector (Kenneth R. French’s HiTec portfolio), where the peak is the highest month-end level in the preceding 24 months. The US market (the whole value-weighted market from the same library) is measured over the same peak-to-trough window. Six such episodes are identified between February 1971 and May 2026. Tech fell harder than the market in three of them (1990, 2000–02, 2021–22); in 1973–74, 1987 and 2007–09 it fell slightly less. HiTec is a stand-in for the Nasdaq’s technology core, not the Nasdaq Composite, whose early decades were dominated by small over-the-counter stocks.

Drawdown comparison: US tech vs US market, 1971–2026

Mean drawdown ratio across all six episodes: 1.22×. Median: 1.15×. Standard deviation: 0.31. Across the three episodes where tech fell harder: mean 1.46×, median 1.33×. All values computed from month-end total-return levels (Kenneth R. French Data Library).

Source: Kenneth R. French Data Library (5 Industry Portfolios, HiTec; Fama-French factors, Mkt-RF + RF), month-end total returns. Chart: Eco3min Research.

A legitimate analytical qualification is that the pattern’s magnitude is driven by a small number of episodes. The 2000–02 (1.73×) and 2021–22 (1.33×) ratios are the two largest in the dataset. Excluding both, the mean across the remaining four episodes is 1.06×, with tech falling harder in 1 of 4. Excluding the two deepest absolute drawdowns (2000–02 and 2007–09) brings the mean to 1.14×. The finding is conditional, and the three ratios close to 1× (0.95× in 1973–74, 0.98× in 1987, 0.99× in 2007–09) are real datapoints, not rounding artefacts.

Real Rates and the Tech/Market Ratio: No Clean 1995 Break

This is the most important honest disclosure of this study. An earlier version of this page, built on index price data, reported a sign reversal around 1995: the ratio moving with real rates before, against them after. The tech/market total-return ratio does not reproduce it: the correlation is mildly negative before 1995 and close to zero after. On 60-month rolling correlations the value changes sign several times between 1993 and 1998, is negative in about 59% of months since 1998, and is positive from late 2002 to 2011 and again since March 2024. What a regime shift does look like in the data is tracked at the scale of the whole economy in the anatomy of macro-financial regimes since the Great Moderation.

A legitimate analytical qualification is that the technology sector’s composition changed in the same years: the rise of software, the IPO surge of 1995–2000, and the entry of mega-cap technology firms. A duration mechanism, if it operates, works through that composition; the weak correlations above suggest it has not dominated the tech/market ratio in either era. If the sector’s composition were to shift again (toward shorter-duration businesses or cash-rich incumbents) its relationship with real rates could change even without a change in monetary policy regime.

Regime Interpretation

What Happened Next? Tech Forward Returns by Real-Rate Regime

This table summarises 12-month forward tech total returns conditional on the real 10-year yield regime at the start of the window. The sample is restricted to post-1995 monthly observations (Jan 1995 onwards through May 2025, the most recent date with a 12-month forward window observable), the modern composition era. The regime classifier is computed in real time from the contemporaneous CPI YoY and 10Y yield, no look-ahead.

Methodological note: Returns are total returns computed from month-end index levels, non-overlapping in the sense that each month is one observation, but the 12-month forward windows do overlap. The classifier uses the contemporaneous ex-post real yield, computed as the 10Y monthly average minus the latest available CPI YoY. Sample is restricted to post-1995 (Jan 1995 onwards) to cover the modern composition era. All three bucket sample sizes are above 70, but the ≥2.5% bucket is dominated by two long expansions (2003–07 and 2023–26) and its median is therefore strongly affected by survivorship of those particular regimes.

Past distributions are not predictive of future outcomes. Regime-conditional statistics describe historical patterns, not expected returns.

Historical Turning Points: Six Episodes That Shape the Pattern

1973–1974: Oil shock and stagflation

Measured with the tech portfolio, 1973–74 is not a case of tech falling harder: from January 1973 to September 1974 the tech portfolio lost 42.5% in total return against 44.9% for the US market (ratio 0.95×), the smallest ratio of the six episodes. The Fed Funds rate rose from 5.94% to 11.34% over the same period as Burns tightened in response to the oil shock. Documentation: the Federal Reserve’s history of FOMC decisions. This episode falls inside the pre-1995 era; the Nasdaq Composite of those years was a small-cap, financial-services-heavy index, a composition that a technology portfolio does not capture.

August–November 1987: The October crash

The October 1987 crash hit tech and the overall market almost equally: from August to November 1987 the tech portfolio lost 29.4% in total return and the market 29.9% (ratio 0.98×). The entire equity complex fell together in October 1987, driven by portfolio-insurance unwind dynamics rather than by a sectoral revaluation. No tightening cycle was active. The episode is consistent with the idea that “tech falls harder” belongs to bear markets driven by valuations, real rates or earnings revisions, not to mechanical position-unwind events. Both drivers belong to the wider mechanics of how monetary regimes and earnings cycles shape equity markets and ETFs.

March 2000 – September 2002: The dot-com bust

The deepest drawdown in the dataset. From March 2000 to September 2002 the tech portfolio lost 77.9% in total return, while the US market lost 44.9%. The ratio is 1.73×, the largest of the six episodes. The Fed Funds rate fell from 5.85% to 1.75% over the same period, yet tech fell harder despite easing. Real rates barely moved over the window (2.49% to 2.35%), so the episode says little about a real-rate channel: it was a collapse of bubble valuations, which compressed faster in tech than in the broader market.

October 2007 – February 2009: The Global Financial Crisis (tech fell as much as the market)

One of the three bear markets where tech did not fall harder than the market. From October 2007 to February 2009 the tech portfolio lost 49.8% in total return and the US market 50.3%. The drawdown ratio is 0.99×. Three mechanisms plausibly explain why tech did not fall harder. First, tech had already corrected 78% in 2000–02 and entered the GFC with relatively low valuations, while the broad market’s exposure to financials (which collapsed) and consumer cyclicals was much greater. Second, the shock originated in the housing and credit complex, not in equity discount rates, affecting the broader market at least as much as tech. Third, the Fed Funds rate fell 454 bp over the window (from 4.76% to 0.22%), an aggressive easing that supported the long-duration tech exposure relative to the immediate cash-flow stress facing financials. The GFC is documented here because intellectual honesty about the limits of the pattern is what makes the pattern worth citing. A related perspective: our analysis of S&P 500 ETF differences.

December 2021 – September 2022: Post-COVID tightening

The most recent identified bear market. From December 2021 to September 2022, the tech portfolio lost 33.1% in total return while the US market lost 24.8%. The ratio is 1.33×. The Fed Funds rate rose 248 bp from 0.08% to 2.56% over the bear-market window, and continued tightening to a cycle peak of 5.33% in August 2023 (a cumulative +525 bp from the Dec 2021 floor). The ex-post real 10-year yield rose from −5.71% to −4.67% over the bear-market window (the rise into positive territory continued through 2023). This is the textbook post-1995 case: real rates rose, the ratio compressed, tech fell harder.

2015–2018: A cycle without a tech bear market

This tightening cycle passed without a tech bear market: the tech portfolio’s peak-to-trough decline did not reach the 20% threshold, so no formal bear market is logged. Over the cycle window (November 2015 – December 2018, +215 bp Fed Funds increase), the tech portfolio fell 16.1% in total return from September to December 2018, against 14.5% for the market over the same months. Tech still fell harder than the market within the cycle window. The cycle is not unique: in 1994–95 the tech portfolio’s largest decline was 6.1% (market 7.6%) and in 2004–06 11.3% (market 4.7%). It is named here to make clear that the pattern is conditional on a bear market being triggered: not every Fed tightening produces one.

May 2026: Current observation

Since its September 2022 trough, the tech portfolio has returned +202.7% (total return, to May 2026). The tech/market ratio (Feb 1971 = 100) is 129, below its February 2000 peak of 149, which it has not regained. The ex-post real 10-year yield is about 0.3% (10-year yield 4.48% minus CPI inflation of 4.17%), in the lower part of the moderate regime band. In the post-1995 sample, months in the 0%–2% band were followed by a median 12-month tech total return of +18.7% (n=162), with substantial dispersion.

Methodology

This dataset combines five primary series into 664 monthly observations spanning February 1971 to May 2026: US technology sector total returns (HiTec portfolio of the 5 Industry Portfolios, value-weighted, Kenneth R. French Data Library), US stock market total returns (Mkt-RF + RF, same library), 10-year Treasury constant-maturity yield (DGS10 via FRED), Federal Funds effective rate (FEDFUNDS via FRED), and Consumer Price Index for All Urban Consumers (CPIAUCSL via BLS, accessed via FRED). HiTec stands in for the Nasdaq’s technology core; Nasdaq Composite and S&P 500 index levels are not used.

Bear-Market and Tightening-Cycle Selection

Sensitivity: lowering the bear-market threshold to 15% adds five episodes (1983–84, 1998, 2011, 2018, 2020), bringing the total to 11, with tech falling harder in 6 of 11 (the exceptions: 1973–74, 1987, 2007–09, 2011 and 2020). Raising the threshold to 25% drops 1990 (tech −20.9%), leaving 5 episodes, with tech falling harder in 2 of 5. The conclusion, tech falls harder in some bear markets but not as a rule, holds across the 15%–25% range.

Real-rate measurement choice

We use ex-post real rates (nominal yield minus contemporaneous CPI YoY) rather than TIPS-implied real rates because the TIPS series begins only in 2003. The ex-post measure gives a uniform 55-year series. For 2003 onward, the correlation between the ex-post and TIPS-implied real 10-year yields is +0.54 in monthly observations (2003–2026); substituting TIPS from 2003 moves the 1996–2026 correlation of the tech/market ratio from −0.08 to +0.08: both values are close to zero. This methodology follows the convention used in Reinhart & Sbrancia (2011, “The Liquidation of Government Debt”).

Dataset Design

Python Reproduction Code

# Reproduce this dataset from primary sources
import pandas as pd
from fredapi import Fred

fred = Fred(api_key='YOUR_FRED_KEY')
# Kenneth R. French Data Library: 5_Industry_Portfolios_CSV.zip (value-weighted, monthly, %) and F-F_Research_Data_Factors_CSV.zip
tech   = (1 + hitec_pct / 100).cumprod()              # HiTec column
market = (1 + (mkt_rf_pct + rf_pct) / 100).cumprod()   # Mkt-RF + RF
dgs10  = fred.get_series('DGS10').resample('M').mean()
cpi    = fred.get_series('CPIAUCSL')
fedfunds = fred.get_series('FEDFUNDS')

cpi_yoy = (cpi / cpi.shift(12) - 1) * 100
real_10y = dgs10 - cpi_yoy
ratio = (tech / market) / (tech.iloc[0] / market.iloc[0]) * 100

# Bear-market identification on the tech index: rolling 24m high; episode = months below it; trough = episode minimum
# Full implementation in /scripts/build_csv.py in dataset bundle
  

Dataset Download & Reproducibility

664 observations · Monthly · February 1971 – May 2026 · 18 columns including bear-market and tightening-cycle identifiers · The study file is not offered for download; the methodology lists every source, so the series can be rebuilt from them.

Data Sources & References

• Primary Kenneth R. French Data Library, 5 Industry Portfolios (HiTec, value-weighted, monthly total returns; file built from the 202608 CRSP database), accessed September 2026.
• Primary Kenneth R. French Data Library, Fama/French 3 Factors (US market return: Mkt-RF + RF, monthly), accessed September 2026.
• Primary Board of Governors of the Federal Reserve System, 10-Year Treasury Constant Maturity Rate (DGS10), accessed via FRED, May 2026.
• Primary Board of Governors of the Federal Reserve System, Federal Funds Effective Rate (FEDFUNDS), accessed via FRED, May 2026.
• Primary U.S. Bureau of Labor Statistics, Consumer Price Index for All Urban Consumers (CPIAUCSL), accessed via FRED, May 2026.
• Research Reinhart, C. M. & Sbrancia, M. B. (2011), “The Liquidation of Government Debt,” NBER Working Paper 16893.
• Research Lettau, M. & Wachter, J. A. (2007), “Why Is Long-Horizon Equity Less Risky? A Duration-Based Explanation of the Value Premium,” Journal of Finance, 62(1), 55–92.
• Reference National Bureau of Economic Research, US Business Cycle Expansions and Contractions.

Methodological Limitations

• The tech portfolio is a sector proxy, not the Nasdaq Composite: it leaves out the Nasdaq’s non-tech listings (small over-the-counter stocks and financials in its early decades) and includes large telecom companies. Its own composition also shifted, from IBM and the telephone companies in the 1970s to software, semiconductors and platforms after 1995, so cross-period comparisons capture composition effects as much as duration effects.
• The dataset does not measure concentration inside the tech portfolio. In a value-weighted sector dominated by a handful of issuers, a shock on three or four names is enough to move the whole, and broad indices that look diversified, at home or across regions, often return to the same few names. Portfolios exposed to tech through several indices can therefore be more correlated than they appear, a dimension the drawdown ratios above do not capture.
• All return calculations use value-weighted total returns (dividends reinvested) as published by the Kenneth R. French Data Library, not index price levels.
• Ex-post real rates use contemporaneous CPI YoY rather than market-implied inflation expectations. The two measures diverge during periods of inflation surprise. TIPS-implied real yields are available only from 2003; correlation with ex-post is +0.54 over the overlap (2003–2026).
• The sample of n=6 tech bear markets is small. The split (tech harder in 3 of 6, about equal in the other 3) holds across the 15–25% threshold range, and no statistical significance against a specific null hypothesis is claimed.
• Forward 12-month return windows overlap. The “% positive” and median figures describe the empirical distribution of outcomes, not independent observations.
• No 1995 break appears in the correlation between the tech/market ratio and real rates; formal structural-break tests (Chow, Bai-Perron) are not applied.

Frequently Asked Questions

Source

Related Eco3min Research

• US CPI Inflation History (1913–2026): How four episodes account for 72% of cumulative purchasing power loss
• Credit Breaks First: HY Spreads Lead Equity Markets, Decomposing how credit signals precede major equity drawdowns
• US Real Interest Rates Dataset (1962–Present): The underlying real-rate series used in this study
• Real Interest Rates vs CAPE Ratio Dataset: The tent-shaped relationship
• 30-Year Treasury Duration Risk: How long-duration Treasuries lost 53% peak-to-trough during 2020–2023
• VIX Volatility Index Dataset: Median 12m S&P 500 return at VIX ≥30: +22.4%