Historical multiples vs current

Quick definition

Historical multiples place current valuations in context by comparing them to past averages. If the S&P 500 trades at 21× earnings and the long-term average is 15–16×, the market is trading at a 30% premium to history. This comparison can identify whether valuations are extreme, but it cannot predict future multiples because the relationship between past and future averages is loose.

Key takeaways

• Long-term P/E average for S&P 500 is 15–16×, with 90% of observations between 10–22×
• Mean reversion is probabilistic, not deterministic: extreme valuations eventually normalize, but timing is unpredictable (can take years or decades)
• Structural changes shift baseline multiples, so blindly reverting to ancient history is naive
• Current is more predictive than distant history: last 20 years' averages matter more than data from the 1970s, when business composition was different
• Valuation percentile is more useful than absolute levels: if current P/E is at 90th percentile historically, contraction risk is elevated regardless of whether the multiple is 20× or 25×
• High historical multiples do not guarantee future returns, but extremely high multiples (>25× for the market) have weak forward return profiles

Historical P/E multiples for the S&P 500

Long-term trend (1926–2024)

The S&P 500's average P/E multiple over nearly a century is 15–16×, with substantial variation:

Period	Median P/E	Range	Notes
1926–1950	12–14×	8–18×	Post-depression lower valuations
1951–1980	13–15×	7–20×	Stable, postwar period
1981–2000	15–18×	10–30×	Rising valuations, "Nifty Fifty" bubble
2001–2010	14–16×	10–18×	Post-tech bust, recovery
2011–2019	18–21×	15–25×	QE stimulus, low rates, expansion
2020–2024	16–21×	12–24×	Pandemic volatility, rate hikes, reset

Current context (2024)

The S&P 500 trades at approximately 20–21× trailing earnings, near the long-term average but above the median of the past 50 years. This is:

• 35% above the 1926–2024 median of 15×
• 5–10% above the last 30-year average of 16–17×
• Below the 2021 peak of 22–23×

By percentile: current valuation is roughly 60–65th percentile historically. Not extreme, but in the upper half of historical ranges.

Sectors and historical valuation norms

Historical multiples differ by sector:

Sector	Long-term avg	Recent avg	2024 level	Status
Technology	25–30×	28–32×	26–28×	Below recent; fair to slightly cheap
Financials	12–14×	10–12×	11–13×	Fair
Healthcare	16–18×	18–20×	19–21×	Slightly expensive
Consumer discretionary	16–18×	18–22×	20–22×	Above long-term; slightly expensive
Industrials	14–16×	15–17×	16–18×	Fair
Utilities	12–14×	13–15×	14–16×	Fair
Energy	10–12×	8–10×	12–14×	Above recent; slightly expensive
Materials	12–14×	13–15×	15–17×	Slightly expensive
Consumer staples	15–18×	18–20×	19–21×	Slightly expensive

Comparisons reveal which sectors are expensive relative to their own history. Financials at 12× is at their long-term average; Technology at 27× is below its long-term average. Context depends on how you define "normal" for each sector.

Why historical multiples compress and expand

Structural shifts in business composition

In 1970, the S&P 500 was heavily weighted toward industrial manufacturing, utilities, and banks. Capital intensity was high; margins low. A 12× P/E reflected this profile.

By 2024, tech and healthcare dominate the market. These businesses are capital-light, margin-rich, and growing faster. A 20× P/E reflects a different asset base. Comparing 1970 and 2024 P/E ratios directly is misleading because the composition of earnings, profitability, and growth profiles changed.

Discount rate changes

When the risk-free rate was 6% (1990s), investors required higher equity returns and accepted lower multiples. When rates fell to 0.5% (2020–2021), required returns fell and multiples expanded. A structural shift in real rates compresses or expands the entire market's base multiple.

Post-2021, risk-free rates rose from 0.5% to 4.5%. This alone justified 25–35% multiple compression. Some compression occurred; not all theoretical compression has realized yet, suggesting further downside risk or that other factors (e.g., lower equity risk premium) offset rate impacts.

Margin expansion

If the S&P 500's average profit margin was 5% in 1990 and 8% in 2024, the same earnings level can support higher multiples because that earnings level is derived from less revenue (more productive capital). Margin expansion justifies higher multiples.

Conversely, if margins compress (labor costs rise, competition intensifies), multiples must contract to prevent overvaluation.

Growth rate changes

Faster growth justifies higher multiples. The S&P 500's real earnings growth was 2–3% annually in the 1970s–1990s and 3–5% in the 2010s (with volatility from cycles). Higher growth supports higher base multiples.

Mean reversion: how strong and how long?

Mean reversion—the tendency of extreme valuations to normalize—is one of the most studied phenomena in investing. The evidence supports it, but with important caveats.

Evidence for mean reversion

Long-term Shiller CAPE data (1880–2024): The cyclically-adjusted P/E (Shiller CAPE) averages 17× historically. When CAPE reaches 30× (1929 peak, 2000 tech bubble, 2021), subsequent 10–15 year real returns have been weak (1–3% annually). When CAPE falls to 8–10× (1949, 1982, 2009), subsequent returns have been strong (10–15% annually).

This is powerful evidence that extreme valuation multiples revert to mean over time.

S&P 500 P/E cycles: When the S&P 500 trades above 20× earnings, subsequent 5-year real returns average 4–7% annually. When it trades below 12×, subsequent 5-year returns average 10–15%. The relationship is inverse: high multiples presage low returns; low multiples presage high returns.

The timing problem

Mean reversion does not tell you when reversion occurs. In 1995, the S&P 500 was trading at 17–18×, well above the long-term average. Investors calling for mean reversion and moving to cash "missed" the subsequent doubling of the market in 1995–2000.

Similarly, valuations remained elevated from 2015–2021. Investors sitting out those years waiting for mean reversion underperformed dramatically. Extreme can persist for years, especially if structural reasons support higher multiples.

Statistical vs. economical mean reversion

Statistical mean reversion is undeniable: valuations regress to the mean over 15–30 year periods.

Economical mean reversion (the assumption that today's mean is the correct mean) is less certain. If the structural cost of capital has fallen, the "new normal" multiple might be 5–10% higher than the historical mean. Reverting to the old mean would mean overshooting downward.

Using historical multiples to make decisions

The percentile approach

Rather than assuming absolute multiples (15× is "normal," 20× is "expensive"), use percentiles. Calculate where current multiples rank historically.

S&P 500 P/E percentile analysis:

• 10th percentile: 10–11× (extreme cheap, rare)
• 25th percentile: 12–13× (cheap)
• 50th percentile: 15–16× (median, neutral)
• 75th percentile: 18–19× (expensive)
• 90th percentile: 21–22× (very expensive)

In 2024 at 20–21×, the market is at roughly 65–75th percentile—in the expensive half of history, but not at extreme. This suggests modest downside valuation risk but no screaming opportunity. Returns should be modest because multiples cannot expand much further; returns depend on earnings growth.

Historical multiples as a valuation floor and ceiling

Rather than a single "fair value" multiple, use historical ranges:

A stock's fair value zone is:

• Floor: 25th percentile of its historical P/E (rarely gets cheaper for this business)
• Fair: 50th percentile (typical valuation)
• Ceiling: 75th percentile (fair but on the expensive side)

If a stock's historical P/E range is 12–20× with a median of 15×:

• Below 12×: consider buying (bargain, 1–2 standard deviations below mean)
• 12–15×: buy (within fair range)
• 15–18×: hold (not expensive, reasonable)
• 18–20×: hold with caution (near historical ceiling)
• Above 20×: consider selling (extreme, structural change must justify it)

Identifying mean reversion opportunities

The most powerful application of historical analysis: identify when valuation extremes appear likely to revert.

Signs of mean reversion risk (downside):

• Current P/E is above 80th percentile historically
• Current P/E is well above sector average
• Sector fundamentals are deteriorating, not improving
• Interest rates are rising, compressing discount rates
• Sentiment is euphoric

Signs of mean reversion opportunity (upside):

• Current P/E is below 20th percentile historically
• Current P/E is well below sector average
• Sector fundamentals are improving, business stabilizing
• Interest rates are falling, expanding discount rates
• Sentiment is pessimistic

Real-world examples

Tech: 2000 bubble vs. 2024

2000 peak: NASDAQ stocks (heavily tech) traded at 50–100× earnings. Historical average for tech was 20–25×. Valuation percentile: 99th. Every signal screamed mean reversion downside.

The Shiller CAPE for the market was 44× (highest in history except 1929). Subsequent 10-year real returns were negative. This was an extreme that reverted savagely.

2024: Tech trades at 26–28× earnings. Historical average for quality tech software is 25–30×. Valuation percentile: 55–65th. No extreme; fairly valued. This is not a mean reversion opportunity, just a normal valuation.

Investors who use "tech was expensive in 2000, therefore expensive now" commit the error of not updating the comparison to account for structural changes (software as a service, recurring revenue, higher margins).

Utilities: 2011 vs. 2024

In 2011, utilities traded at 12× earnings; historical average was 13–14×. Below median. Sentiment was depressed from post-2008 malaise. This was a classic mean reversion opportunity.

Utilities trades at 14–16× in 2024, reflecting structural changes: rising capital needs for grid modernization and renewables, regulatory pressure on returns, rising rates. The "normal" multiple for utilities is probably 13–14× now (slightly lower than 2011 because of structural headwinds). 2011 offered a 15–20% upside from mean reversion; 2024 offers little.

Regional Banks: 2023 opportunity

After the March 2023 regional bank crisis (SVB collapse), many regional banks traded at 0.5–0.7× book value and 6–7× earnings. Historically, healthy regional banks trade at 0.9–1.2× book and 10–12× earnings.

The mean reversion upside from 2023 lows was substantial—50% or more if banks normalized. This was a classic opportunity: extreme compression created by panic, not fundamental deterioration (most banks were solvent, just illiquid).

Over the next year, banks partially recovered. Those that bought in 2023 at extremes outperformed those who waited.

Common mistakes in historical analysis

Mistake 1: Treating historical average as the "true" fair value

The 15–16× P/E average for the S&P 500 is a historical observation, not a law of nature. If business composition, margins, and growth rates permanently shift upward, the "new normal" might be 18–20×. Reverting to 15× would be overshooting.

Mistake 2: Using too much history

Data from 1950 is interesting but less relevant than data from 2000. Business models, competition, capital structure, and sector composition have changed. Weight recent history more heavily. Use the last 20–30 years as your "normal," not the last 100.

Mistake 3: Ignoring dividend changes

A company that paid 4% dividend in 1990 and 2% in 2024 is not at the same valuation on a total-return basis even if P/E is identical. Adjust for dividend yield changes when comparing historical valuations.

Mistake 4: Confusing percentile with prediction

Being at the 80th percentile of valuation does not mean the stock will decline to the 50th percentile next year. It means it is more likely to do so eventually, but timing is uncertain. Do not overweight percentile signals alone.

Mistake 5: Not adjusting for interest rates

In a 5% rate environment, equities should trade at lower multiples than in a 1% environment, all else equal. A 15× P/E in 2024 (rates at 4.5%) is not directly comparable to a 15× P/E in 2020 (rates near zero). Adjust for rate changes when using historical comparisons.

FAQ

What is the Shiller CAPE ratio and why does it matter?

The Cyclically-Adjusted P/E (CAPE) adjusts earnings for the business cycle by using a 10-year average earnings rather than trailing earnings. This smooths out cyclical extremes. High CAPE (>30) has predicted weak future returns; low CAPE (<10) has predicted strong returns. It is more predictive of long-term returns than standard P/E.

If the S&P 500 is at the 65th percentile historically, should I be bearish?

Not necessarily. Being in the upper half of history is not inherently bad. If interest rates are lower than historical average, growth is higher, or margins are superior, the valuation is justified. Use percentile as one signal, not the entire framework. Combine with absolute valuation (DCF) and trend analysis (improving vs. deteriorating fundamentals).

How long does mean reversion typically take?

Evidence suggests 10–15 years for extreme valuations (>25× or <10×) to revert. But some extremes persist for decades if structural changes justify them. The dot-com bubble reverted in 2000–2010; the low valuations of 1982 expanded for years thereafter. Expect reversion, but do not time it to the year.

Should I avoid stocks at higher-than-historical multiples?

Not automatically. If the business is genuinely improving (margin expansion, growth acceleration, lower risk), higher multiples are justified. IBM at 20× in 2010 was expensive relative to history but justified by stable dividends and buybacks. Context matters more than absolute level.

What sectors are currently cheap based on historical multiples?

This varies by date. As of 2024, Financials and Energy trade slightly below their long-term averages; Tech trades near its long-term average; Healthcare and Discretionary trade above. But "cheap" must be paired with whether the business is improving or deteriorating, not just whether valuation is below historical average.

Can I use historical multiples for a stock I'm analyzing for the first time?

Yes. Use the company's own 10–15 year historical range if available. If not, use sector peers' historical ranges as proxy. A software company with no long history can be evaluated against the software sector's historical norm and against direct peers.

Related concepts

• Shiller CAPE ratio — cyclically-adjusted P/E, more predictive of long-term returns than standard P/E
• Discount rate and required returns — the mechanism by which interest rates change appropriate multiples
• Business cycle and valuation timing — how economic cycles create valuation extremes
• Earnings growth and margins — the fundamental drivers that can justify structural multiple changes
• Multiple expansion and contraction — the dynamics of how multiples move around historical averages
• Mean reversion — the statistical and economical tendency of extremes to normalize

Summary

Historical multiples provide critical context for current valuations. The S&P 500's long-term P/E average is 15–16×; when it trades significantly above or below this, mean reversion risk rises. However, structural changes can shift the baseline, so blindly reverting to distant history is naive. Use historical percentiles to identify where valuations rank (10th to 90th percentile); extremes above 80th or below 20th percentile are more likely to revert. Recent history (20–30 years) matters more than distant history. Combine historical analysis with absolute valuation and fundamental analysis: a stock trading at an expensive percentile can still be attractive if fundamentals are improving and the business supports higher multiples. Mean reversion is powerful over 10–15 years but can be frustratingly slow or delayed; do not overweight timing predictions. The strongest setup: valuation at percentile extreme (high or low) plus fundamentals in line with that valuation moving further or confirming the mean reversion opportunity.

Next

Read Cross-industry multiple comparisons to learn how multiples vary by industry and why comparing multiples across sectors requires careful adjustment for fundamental differences.