Are 2005-era margin-of-safety thresholds still the right defaults? A backtest, a recency-bias check, and a vivid mean-reversion scenario.
Research compiled May 15, 2026 — Flint / Coleman Research
Drag the macro sliders on the left to model the structural regime. Drag the framework sliders on the right to test where you want your floors set. The verdict box reclassifies in real time.
Sources: multpl.com (annual averages); Shiller dataset. Horizontal lines show 1985–2005 vs 2005–2026 means.
Sources: multpl.com; currentmarketvaluation.com. Pre-2000 average ~1.5; post-2010 average ~2.3; current 3.67.
FRED M2SL. Vertical annotations: QE1 Nov 2008, QE2 Nov 2010, QE3 Sept 2012, COVID QE Mar 2020, QT start Jun 2022.
Peak-to-trough %. 50% MOS only triggered in 2000–2002 and 2008–2009.
Apollo, Yale/Chicago, GWU sources. Index funds went from ~7% (2000) to ~33% (2024).
Long-run pre-2000 mean ≈16.3. Post-2010 mean ≈31.8. The market is structurally re-rated ~2x.
One floor doesn't fit all. A floor that's tight for Costco is loose for U.S. Steel. The framework below differentiates by gross margin, capital intensity, and durability of cash flow — the three variables that actually determine what multiple a business deserves.
Microsoft, ServiceNow, Adobe, Salesforce, mid-cap pure-play SaaS. These businesses have 75-85% gross margins (chargebee, cloudzero benchmarks), capital-light scaling, and recurring revenue. A 30%+ FCF margin and 80%+ gross retention means $1 of incremental revenue is worth ~5x more than $1 at a 25%-margin industrial. Phil Town's 15x P/E rule, applied here, would have made you skip every meaningful winner of the last 15 years.
Visa, Mastercard, Moody's, S&P Global, ICE, CME, MSCI, FactSet. Transaction-based or licensing revenue with near-100% incremental margins. The P/S ratio here looks insane on Phil Town's chart but the businesses generate cash like a tax. Buffett's See's Candy framework applies — pay up because the reinvestment economics are extraordinary. Hard ceiling matters: when V/MA cross 35x P/E with sub-10% growth, that's expensive even on this scale.
Costco, Walmart, Chipotle, McDonald's, Nike (when on sale), Starbucks. Lower gross margins than software but durable brand, repeat demand, predictable unit growth. Costco trades at 52x P/E with a fair-value estimate near $295 vs $1,011 share price (ValueInvesting.io, May 2026). The market has accepted that for a decade and you outperformed by ignoring the cheap screen. This is the tier where Phil Town's 15x rule was most damaging — it made you sell Costco at $50 in 2010 and miss a 20-bagger.
Caterpillar, Deere, Honeywell, Parker Hannifin, Eaton, Rockwell. Solid ROIC (15%+) but cyclical demand. Phil Town's rules approximately work here but with a bit more cushion. Use mid-cycle earnings for the P/E, not peak.
U.S. Steel, Nucor, integrated oil & gas, refiners, basic materials, REITs. Here Phil Town's 15x/50% rules basically still apply. These are the businesses he was actually pricing for in 2006: Whirlpool, Energizer, Harley-Davidson — and the rules work because the business economics haven't structurally improved. Cyclical companies need cycle-adjusted earnings, not peak.
Coal, offshore drillers, ocean shippers, leveraged metals, regional bank stress trades. Phil Town floors with an extra 10% MOS cushion because the downside isn't just multiple compression — it's actual impairment of the business. Don't use trailing earnings; use a normalized through-cycle figure.
The investor who applies tiered floors wins in both regimes. In a continued-passive regime, they own Tier 1-3 compounders that benefit from the ETF bid. In a reversion regime, the tighter Tier 5-6 floors keep cyclical positions from getting massacred because they were bought with 40-50% MOS. The investor who applies uniform Phil Town floors across all sectors misses every Tier 1-3 winner from 2010-2025 and ends up in nothing but Tier 5-6 deep value, which underperforms in QE regimes and gets killed in shocks because the businesses are structurally weaker.
Phil Town's Rule #1 (2006) embedded three numerical floors that have lived inside Flint's stock-research skills for almost two decades: maximum 15x P/E, maximum 2x sticker price (effectively 2x P/S for many businesses), minimum 50% margin of safety. Those floors were calibrated against an investment universe that no longer exists. Three structural forces — passive/401k inflows (~33% of US equity AUM versus near-zero in 1985), the 15-year ZIRP/QE regime (Fed balance sheet went from sub-$1T to $9T+), and the rise of software economics (75% gross margins versus 30% for the consumer-staples that Phil Town modeled) — have re-rated equity multiples roughly 2x at the index level. The Shiller CAPE averaged 16.3 pre-2000 and 31.8 post-2010. S&P P/S averaged 1.5 in 1985-2005 and is 3.67 today. These are not noise; they are the new mean. The right action is not to abandon margin of safety but to retune defaults to roughly 4x P/S, 22x P/E, 25% MOS at the framework level, then tier by business type. The strict Phil Town floors stay alive but only for capital-heavy, commodity-cyclical businesses — the kind Phil Town was actually pricing for. Recency bias is a real counter-argument; the honest probability of full mean reversion to 1985-2000 levels is maybe 10-15% on a 10-year horizon, and the mean-reversion path is non-linear and devastating when it happens.
Total US retirement assets reached $49.1T at year-end 2025 (ICI Fact Book 2026), with $14.2T in DC plans and 401(k)s holding $10.1T alone. Index funds and ETFs now hold roughly 30-34% of the US equity market (Apollo Academy, GWU, Yale/Chicago studies), up from under 10% in 2000. Mike Green's thesis at Simplify: passive inflows are mechanical — they buy the index in proportion to weights without any reference to price. This creates a feedback loop: capital flows in → biggest stocks get biggest share of the bid → biggest stocks get more expensive → their index weight increases → they get more flow. Multiple expansion in mega-caps is a partly mechanical consequence, not a fundamental judgment.
This is durable as long as 401k contributions exceed retirement withdrawals — which on demographic projections holds until roughly 2035, when Boomer drawdowns may finally exceed Gen X / Millennial / Gen Z contributions. But "may" is doing real work; nobody knows the exact cross-over because rebalancing inside retirement accounts also matters.
The Fed balance sheet went from $0.9T in 2008 to a peak of $8.9T in 2022. M2 money supply (FRED M2SL) went from $2.5T at end of 1985 to $22.4T at end of 2025 — a 9x increase. QE programs: QE1 (Nov 25, 2008, $600B agency/MBS), QE2 (Nov 3, 2010, $600B Treasuries), QE3 (Sep 13, 2012, $40B/month open-ended), COVID QE (Mar 15, 2020, initial $700B, expanded to unlimited). The math is clear: discount rates fell, so terminal multiples rose. Damodaran's 2025 update shows implied US ERP at 4.33% with an implied cost of equity around 8.91% — historically reasonable, but combined with a 4.18-4.46% Treasury yield, the discount rate complex is still well below the 1985-2005 average.
The math: terminal multiple ≈ 1/(r - g). If discount rate r drops from 11% to 8% and growth g is 3%, terminal multiple goes from 12.5x to 20x — a 60% expansion just from rates. That's roughly the level of multiple expansion the market saw 1985 to 2025.
SaaS companies operate at 70-85% gross margins (Chargebee/cloudzero benchmarks: 75% is the industry standard, best-in-class 80%+). Traditional consumer goods are 30-40%. Heavy industrials are 20-30%. The fundamental valuation math says you should pay a higher multiple of revenue for a business that converts more of each revenue dollar to gross profit, free cash flow, and reinvestable capital. Microsoft's gross margin is ~70%; Whirlpool's is ~16%. Paying 5x P/S for MSFT is roughly equivalent in cash-generation terms to paying 1.5x P/S for Whirlpool. The P/S floor of 2x is not wrong — it just doesn't apply to the same universe.
Phil Town's Rule #1 framework:
His 2010 follow-up Payback Time didn't change the floors. He moved emphasis toward "quality compounders" via the "4M" framework (meaningful moat, growth, management, margin of safety), and introduced "payback time" (years to recoup price via FCF) as a complementary risk filter, but kept the 50% MOS and the 15x P/E rule of thumb.
Here's what Phil Town genuinely got right that the recalibration preserves: require a margin of safety, do not pay for businesses you don't understand, treat the price you pay as a determinant of long-term returns. The floors are the disposable part of the framework; the discipline of demanding a discount to intrinsic value is the durable part.
Buffett's 1999 Fortune piece (Berkshire PDF) is the most precise primary source. The key quotes:
"What is important is whether the ratio of total stock market value to GDP is at a normal level."
"If government interest rates, now at a level of about 6%, were to fall to 3%, that factor alone would come close to doubling the value of common stocks."
That second quote is structurally important. Buffett himself acknowledged in 1999 that halving the discount rate roughly doubles equity values — exactly the mechanism that ZIRP/QE created over the following two decades. The "Buffett indicator" (market cap to GDP) is now around 230-232% versus his original "playing with fire" threshold of 200% (longtermtrends, currentmarketvaluation). He hasn't repeated the explicit 200% line at recent annual meetings but consistently said in 2020-2022 that stocks are "not cheap." In 2024 he built Berkshire's cash position to $325.2B while trimming Apple — a more honest signal than any quote.
The Munger/See's Candy framework matters too. In 1972, Buffett paid 3x book and ~11x earnings for See's — multiples that would have screamed "expensive" to a strict Graham cigar-butt investor. The investment generated $1.35B in pre-tax earnings on $32M of incremental capital over 30+ years. Munger's lesson: "A great business at a fair price is superior to a fair business at a great price." The whole 2006 Phil Town framework is a Graham-style cigar-butt rule. The recalibration moves it toward Munger.
Pal's framework treats global liquidity as the master variable driving asset prices since 2008. Key claims:
Where Pal's framework holds: the empirical correlation between global liquidity growth and risk-asset prices since 2009 is real and well-documented (Druckenmiller has said the same thing for 40 years: "earnings don't move the overall market; it's the Federal Reserve Board"). The 401k/passive-flow mechanism Pal points to overlaps with Mike Green's work.
Where Pal is just-so storytelling: the "Everything Code" framing implies almost deterministic forward predictability ("$100 trillion crypto") which is not how macro actually works. The exact mapping between Bitcoin halvings and US debt refinancing cycles is loose-fit pattern-matching. The 8% debasement figure is rhetorical — there's no clean way to measure it. Treat Pal's framework as a directional lens (liquidity matters, debt rollover matters, multiples respond to both) but not a quantitative prediction engine.
Imagine the structural drivers reverse. Demographics: Boomer drawdowns finally exceed Gen X/Millennial 401k contributions in 2031, accelerated by inflation forcing earlier-than-planned retirements. Rates: a second wave of inflation in 2027-2028 pushes the 10-year back to 6.5-7.5% sustained. Passive flows turn net negative as forced sellers in retirement accounts dominate forced buyers.
This is what that world looks like in concrete terms:
The mechanics: Vanguard, BlackRock, Fidelity 401k platforms see net outflows for the first time in modern history. The mechanical bid disappears. Stocks that benefit most from the passive bid — concentrated mega-caps with high index weights — sell off hardest. Magnificent Seven names compress from 30-40x P/E toward 15-20x. The S&P 500 multiple compresses from 22x forward to 14x. That's a 36% multiple compression even without earnings declining. If earnings also compress 20% in the recession, you get a 50% drawdown — the kind not seen since 2008-2009.
Where the money goes: The first wave goes to Treasuries as the safe-haven trade. The 10-year crashes from 7% to 4% in six months. Gold rips because central banks are forced to ease into the recession, debasing currency. Cash equivalents stay attractive because front-end yields lag the recession-cut path. Real estate bottoms slow because of rate-pinned mortgages. Crypto initially crashes 60-70% in the liquidity squeeze, then ramps as the next QE wave hits. International / emerging markets outperform because US-specific valuation excess unwinds. GMO's 7-year forecast (May 31, 2025) calls this scenario: US Large Cap -4.9% real return, US Small Cap -2.9%, International Large Cap +2.8%, EM +5.2%, US Bonds +1.8%.
What people are doing: A Boomer who was supposed to retire at 67 with $2.4M in their 401(k) sees the account drop to $1.4M in eight months. They delay retirement by 4-7 years. Many freeze and don't sell; the panicked subset sells at the bottom, locking losses. Hardship withdrawals from 401(k)s spike as the recession lays off Gen X workers in their peak earning years. Younger workers cut 401(k) contributions to maintain take-home pay. Allianz Center for the Future of Retirement: 67% of Americans already worry more about exhausting savings than death — that fear becomes operative reality.
The compounded effect: Forced selling from delayed retirements + reduced 401(k) contributions + Fed cuts that don't restore confidence quickly + commercial real estate workout + regional bank stress = a 24-36 month bear market with multiple compression as the primary driver. The S&P drops from 6,300 to ~3,500 peak-to-trough. Hussman's MarketCap/GVA mean-reversion call gets vindicated: he's been forecasting a 70%+ drawdown for years and on this path he'd finally be right (though probably ~50%, not 70%).
Generational wealth destruction: The cohort that bought the top in 2024-2025 (peak passive flow) takes the brunt. The cohort that was in cash or short-duration Treasuries (Buffett's posture) buys the bottom. Phil Town's strict 50% MOS rule becomes actively useful for the first time in 20 years — there are real deep-value names trading at 8x earnings with 4% dividend yields and clean balance sheets. The investor who maintained tiered floors with embedded MOS buys aggressively here. The investor who applied uniform Phil Town floors all along has been mostly in cash for 15 years (because nothing met the rules) and now has dry powder. The investor who threw out floors entirely in the bull market gets crushed.
What stays expensive: Genuinely durable compounders with growing earnings — Costco, MSFT, MA, V, MCO. They compress 25-30% but don't crash. Their underlying earnings grow through the recession (Costco grew through 2008-2009). On the other side, Tier 5-6 stocks (the steel, banks, basic materials) get destroyed first, then become the gen-defining buys.
This is the strongest objection and worth steelmanning carefully. The 1929-1932 drawdown was 86%. The 1973-1974 bear was 48%. The 2000-2002 bear was 49%. The 2008-2009 bear was 57%. In each case, multiples compressed back toward long-run means. From 1929-1949 the Shiller CAPE averaged ~12. From 1973-1985 it averaged ~10. The post-2010 average of 31.8 is genuinely historic — not just elevated, but a regime never seen before in 145 years of data.
The honest response: the recalibration does assume the post-2010 regime is durable for the next 10-15 years. If you put 100% weight on long-run history, you'd keep Phil Town's floors and miss compounders. If you put 100% weight on the last 15 years, you'd abandon MOS entirely and get killed in any shock. The recalibration is roughly 70% weight on post-2010 regime, 30% weight on long-run mean reversion. That blend is defensible but not certain. My honest probability estimate: 10-15% chance of full reversion to pre-2000 norms (CAPE ~16) on a 10-year horizon, 40% chance of a meaningful but partial reversion (CAPE 22-26) on the same horizon, 45% chance the current regime substantially persists.
The academic literature is more cautious than financial-media commentary. The Brookings review of QE transmission channels and CBO analysis both find QE lowered long-term yields through signaling, portfolio balance, and safe-asset channels. But permanent re-rating of P/E ratios is not established; FPA's review of academic studies says QE1/QE2 had "little discernable effect on equity valuations" while later programs had measurable impacts. The truthful answer: QE/ZIRP demonstrably lowered discount rates, which mechanically raises multiples; whether that's "permanent" depends on whether rates go back up, and Howard Marks' "sea change" thesis is that they mostly won't (he expects 2-4% as the new floor, not 0-2%).
The recalibration is robust to this counter: even if you assume QE/ZIRP added "only" 5-7 P/E points (not 10-15), the new framework's 22x default is still consistent with a normalized rate environment around 3.5-4.5%. The framework breaks if rates spend the next decade at 6-7%+ — in which case Phil Town floors look genuinely smart again.
This is real and the timing is the only question. Net 401(k) outflows from Boomer retirements probably begin in earnest 2030-2035. Net immigration policy, fertility trends, and Gen Z 401(k) participation rates all push the cross-over later than naive demographics suggest. But the recalibration must include a 25-30% probability of meaningful passive-flow reversal within 10 years. That probability is high enough to justify keeping Phil Town's strict floors alive for Tier 5-6 (capital-heavy/cyclical) where the businesses themselves carry recession risk, and to maintain a tighter MOS in Tier 1-4.
Each of those critics has been "wrong on timing, right on direction." Hussman's MarketCap/GVA has flagged extreme overvaluation since 2014 and the market doubled anyway. Grantham's 7-year forecast for US Large Cap has been negative real return for most of the last 10 years and US Large Cap returned ~12% real annualized. Shiller's CAPE warning has been operative since 2015.
But — and this is important — being wrong on timing for 15 years is exactly what you'd expect in a regime change. Their process is fine; their frame assumes mean reversion to historical norms that the structural forces have suppressed. The recalibration treats them as late-cycle warning signs, not real-time signals. When passive flows turn or rates spike sustainably, those frameworks will become accurate quickly.
All data as of May 15, 2026. Ratings: , , , , .
Prices are approximate; verify before trading. The point of this table isn't precise target prices — it's showing where each name sits on the old vs. new floor framework.
| Ticker | Name | Tier | Approx P/E | Approx P/S | Old Floor? | New Floor? | Rating |
|---|---|---|---|---|---|---|---|
| NUE | Nucor | 5 | ~12x | ~1.4x | PASS | PASS | BUY |
| MET | MetLife | 5 | ~10x | ~0.8x | PASS | PASS | STRONG BUY |
| KEY | KeyCorp | 5/6 | ~9x fwd | ~1.5x | PASS | PASS | BUY |
| COF | Capital One | 5 | ~12x fwd | ~1.6x | PASS | PASS | BUY |
| MOS | Mosaic | 6 | ~15x | ~1.2x | PASS | PASS | BUY |
| BRK.B | Berkshire Hathaway | 3 | ~21x fwd | ~2.3x | FAIL | PASS | STRONG BUY |
| COST | Costco | 3 | ~52x | ~1.4x | FAIL | TIGHT | HOLD |
| MSFT | Microsoft | 1 | ~33x | ~13x | FAIL | PASS | BUY |
| MA | Mastercard | 2 | ~35x | ~17x | FAIL | TIGHT | HOLD |
| V | Visa | 2 | ~32x | ~16x | FAIL | TIGHT | HOLD |
| ICE | Intercontinental Exchange | 2 | ~26x | ~10x | FAIL | PASS | BUY |
| CMG | Chipotle | 3 | ~30x | ~5x | FAIL | TIGHT | HOLD |
| GPN | Global Payments | 2/3 | ~12x fwd | ~2.2x | PASS | PASS | BUY |
| EG | Everest Group | 5 | ~9x | ~1.0x | PASS | PASS | BUY |
Berkshire is the perfect bridge between old and new frameworks: Buffett's $325B cash pile is the cleanest signal that the operator who literally wrote the playbook is positioning defensively. Trades at ~21x forward earnings with $325B cash optionality and a portfolio of compounders. MET passes both old and new floors and benefits from rate normalization.
Mix of strict-floor deep value (NUE, KEY, COF, MOS, EG, GPN) and Tier 1-2 compounders trading at reasonable Tier prices (MSFT, ICE). The deep-value names are insurance against mean reversion. The compounders provide upside in the continued-regime scenario.
All four are world-class businesses that are priced for perfection even on the new framework. COST at 52x P/E is genuinely tight even allowing for premium multiples. MA and V at 32-35x P/E and 16-17x P/S are at the top of Tier 2 floors. CMG at 30x P/E for slowing same-store-sales growth is bordering on uncomfortable. Hold if owned; don't initiate at these prices.
This artifact recalibrates the value-floor defaults inside stock-researchmainstreet-research
For 1985-2005 versus 2005-2025 averages, I used multpl.com annual averages for S&P 500 trailing P/E. From the queries: 1985 P/E was 10.3; 1999 peak was 35.5; 2005 was 23.5; 2025-26 is 32.04 (May 2026 reading). Pre-2000 long-run mean for Shiller CAPE was ~16.3; 2010-2025 average is ~31.8. S&P P/S pre-2000 averaged ~1.5; post-2010 ~2.3; current 3.67.
The structural break in the data is real. The post-2005 era is fundamentally different. The honest read: 2005-2025 multiples were elevated relative to long history but consistent relative to the three structural forces operating in that window. If you believe those forces persist, the new defaults are correct. If you believe they reverse, Phil Town's old defaults reactivate.
The model computes an "implied floor" given the macro regime sliders, then compares it to where you've set the framework sliders manually. Verdict bands:
stock-research
--conservative flag that reverts to Phil Town 2006 defaults (2x/15x/50%)mainstreet-research
FLOOR_REGIME=current|conservative|aggressive so the user sees which framework was applied--conservative flag activates by default. Tier 5-6 deep value becomes the primary opportunity set.Compiled from 50 Perplexity queries run May 15, 2026. Only verified-from-research URLs included — no fabricated links.