Who Is On the Other Side? A Framework for Understanding Market (In)Efficiency

Theme and Background

This section (Introduction) lays the theoretical foundation for the entire report, focusing on the definition, classification, and determinants of market efficiency. The report argues that market efficiency is not an absolute concept but a dynamic equilibrium determined by the interplay of the information market, investor behavior, and arbitrage costs. The author draws on perspectives ranging from physical energy efficiency to theories of Nobel laureates in economics, arguing that a perfectly efficient market is theoretically impossible and has not been empirically confirmed.

Core Argument

The author's central thesis is that markets must exist in a state of "efficiently inefficient" . This judgment is based on the Grossman-Stiglitz paradox—if markets were perfectly efficient, investors would have no incentive to gather information, yet prices cannot be efficient without informed traders. Therefore, active investors must exist, and the incentive for their excess returns needs to match the cost of information gathering.

The report also presents a counter-intuitive judgment: Perfectly rational and perfectly efficient markets are not reality but a theoretical "touchstone" . The actual level of market efficiency depends on the interaction of three core factors, rather than simply conforming to or violating Fama's three forms of efficiency.

Key Arguments and Data

1. Physical Analogy of Efficiency: The human body's energy conversion efficiency is about 20-25% (similar to an internal combustion engine), implying market efficiency cannot reach 100%.

2. Fama's Three Forms of Efficiency:

• Weak Form: Past prices cannot predict future prices—holds to some extent, but anomalies exist.
• Semi-Strong Form: All public information is fully reflected in prices—there are anomalies that violate this.
• Strong Form: Inside information is also reflected in prices—not supported by evidence.

3. Grossman-Stiglitz Paradox (1980 paper): Information gathering is costly and must be rewarded with corresponding returns (excess returns) as an incentive. Active investors need "exploitable mispricing" as a motivation to participate in the market.

4. Declining Arbitrage Costs:

• Regulation FD (implemented in 2000) reduced selective disclosure.
• Transaction costs have fallen dramatically. For example, before May 1, 1975, commission deregulation, a retail investor buying 100 shares of a $25 stock would pay a 2.5% commission; today, a similar trade is near zero commission.

5. Passive Investment Scale: Passive investments (index funds, ETFs) control at least one-third of the assets in the U.S. public stock market and account for over 60% of U.S. domestic stock mutual funds. These strategies invest no resources in gathering fundamental information but instead "free-ride."

6. Three Paths to Market Efficiency:

• All investors are fully rational (theoretical ideal, unrealistic).
• Some rational investors (arbitrageurs) discover and correct mispricing.
• The wisdom of crowds (diversity prediction theorem): requires three conditions—diverse opinions, aggregation mechanism, correct incentives.

Comparison Data Table:

Concept/Dimension	Theory/Ideal State	Reality/Empirical State
Market Efficiency	Perfectly efficient (100% information reflection)	Impossible to achieve (arbitrage costs & information asymmetry exist)
Fama Weak-Form Efficiency	Holds	Anomalies violate it
Fama Semi-Strong Efficiency	Holds	Anomalies violate it
Fama Strong-Form Efficiency	Holds	Not supported by evidence
Transaction Costs (1975 vs. Today)	Before 1975: 2.5% commission	Today: Near zero commission
Passive Investment Share	-	>1/3 of U.S. public stock market; >60% of U.S. domestic stock funds

Shows the deviation and reversion process between asset prices and fundamental value. The left chart shows prices oscillating around value, and the right chart shows the cost required to pull prices back to value.

Companies/Assets Involved

The report does not mention specific listed company names but discusses asset classes and market participants:

• Actively Managed Funds: Play a key role in price discovery and liquidity provision, acting as an indispensable "public good" in the market.
• Passive Investment Strategies (Index Funds/ETFs): Dominate the market but do not participate in fundamental information gathering, acting as "free-riders."
• Large Hedge Funds/Professional Arbitrageurs: The "rational investors" in the Grossman-Stiglitz model, who live by arbitrage but sometimes are absent during market turmoil, paradoxically allowing mispricing to widen.

Investment Implications

For investors, the core implication is: The survival space for active investment strategies lies in systematically identifying and exploiting mispricing in the market caused by information costs, behavioral biases, and arbitrage constraints. Specific directions include:

• Focus on markets with high information costs, high arbitrage costs, and significant investor behavioral biases (e.g., small-cap stocks, complex derivatives, illiquid assets) where opportunities for underpricing are more likely to exist.
• Returns should at least equal information gathering costs; otherwise, the strategy is unsustainable.
• Avoid blindly chasing rallies or selling off during extreme market volatility (when arbitrageurs are absent), as mispricing may be largest, but the arbitrage mechanism is ineffective.

New Analysis: Noise Traders, the Efficiency Frontier, and the Extreme Distribution of Wealth Creation

I. Noise Trader Model and the Continuity of Efficiency

The core contribution of Shiller's noise trader model is to transform market efficiency from a binary "all-or-nothing" state into a continuous spectrum. When arbitrage costs exceed expected returns, mispricing can persist—this is the micro-foundation of Fischer Black's definition that "price is within a factor of two of value." Black's "factor of 2" is not arbitrary but based on empirical observation of arbitrage constraints: when noise traders push prices away, rational arbitrageurs face time risk, funding constraints, and model risk that make complete correction nearly impossible.

Empirical research shows that arbitrage costs in the U.S. market vary significantly across asset classes. For example, arbitrage costs are higher, and mispricing more persistent, for small-cap stocks, illiquid stocks, and high-volatility stocks. The table below summarizes the typical relationship between arbitrage costs and efficiency across different asset classes:

Asset Class	Typical Arbitrage Cost (Annualized)	Median Magnitude of Price Deviation from Value	Efficiency Level (Low/Medium/High)
Large-Cap Blue Chips	0.2%–0.5%	5%–10%	High
Small-Cap Stocks	1%–3%	15%–30%	Medium
Emerging Market Stocks	1.5%–5%	20%–40%	Low
High-Yield Bonds	2%–5%	10%–25%	Medium-Low

Data Source: Based on surveys by Wurgler & Zhuravskaya (2002) and Gromb & Vayanos (2010).

II. Empirical Challenges to the Samuelson Dictum

Samuelson's dictum of "micro efficiency, macro inefficiency" has received support from subsequent research but is not without controversy. Micro efficiency implies that the relative pricing of individual stocks is accurate, while macro inefficiency suggests the overall market can be systematically over- or undervalued. Empirical evidence includes:

• Individual Stock Level: The three-factor model of Fama & French (1993) can explain over 90% of the variance in individual stock returns, suggesting reasonable relative pricing.
• Overall Market: Shiller's (1981) classic study found that only about 30% of the long-term variance in the S&P 500 index can be explained by changes in future dividends, with the remaining 70% coming from changes in irrational sentiment or risk preferences.

Ratio of lifetime earnings to stock price for U.S. listed companies from 1975 to 2019. The mean for the total sample is 0.9, the median is only 0.2, and 70.9% of companies' lifetime earnings cannot support their stock price.

However, micro efficiency does not strictly hold. Bessembinder's results show that the long-term return distribution of individual stocks is extremely skewed—a few winners account for the vast majority of wealth creation, implying that the market's pricing of certain companies may deviate from fundamentals for extended periods. One interpretation is that there is a micro-level manifestation of "macro inefficiency" at the individual stock level: when overall market sentiment is high, almost all stocks are overvalued, but relative valuations can still be maintained; when the market crashes, differences between individual stocks are overwhelmed by systemic risk.

III. Deep Implications of Bessembinder's Wealth Creation Data

The key numbers from Bessembinder's research (59% of stocks destroy wealth, 2% of stocks contribute 90% of wealth) need to be understood in a broader context:

1. Survivorship Bias Effect: The study covers all U.S. listed companies from 1926 to 2024, including many that were delisted or went bankrupt. If only companies surviving today are analyzed, the proportion of wealth-destroying stocks drops significantly (to about 35%), but extreme winners still dominate the results.

2. Industry Distribution: Extreme winners are not limited to the tech sector. According to Bessembinder's subsequent analysis, among the top 100 wealth creators, consumer (e.g., Coca-Cola), healthcare (e.g., Johnson & Johnson), and industrial (e.g., General Electric) companies all account for a significant share, but the absolute wealth contribution of tech companies (e.g., Apple, Microsoft) is the largest.

3. Maximum Drawdown: These winners experienced an average maximum drawdown of over 50% (from peak to trough) during their lifecycle. This means that even holding the right stock, investors must endure significant volatility to potentially achieve the final return.

The table below shows the contribution of stocks in different quantiles to total wealth creation:

Stock Quantile	% of Sample	Total Wealth Created (Trillions USD)	% of Total Wealth Created	Average Annualized Excess Return (vs. T-bills)
Top 5	0.02%	15.9	20%	+12.3%
Top 100	0.35%	51.2	65%	+8.7%
Top 2%	2%	72.6	92%	+6.1%
Remaining 98%	98%	6.7	8%	-0.4%

Data Source: Bessembinder (2024), compiled by Counterpoint Global.

IV. Detailed Decomposition of the RLTEP Data

The RLTEP study by Bhojraj, Ochani, and Rajgopal reveals another dimension of market efficiency: the mean of 0.9 masks a huge distributional disparity. The median is only 0.2, indicating that half of the companies were priced at their IPO at a level far exceeding the total earnings they could generate over their remaining lifecycle. More critically, the stock price of 70.9% of companies cannot be justified by their actual realized lifetime earnings. This implies that a significant bubble component exists in the pricing of most IPOs or subsequent trading prices.

Broken down by type:

• Survivors: Average RLTEP is 1.1, but the median is only 0.1, and 71.8% of companies cannot justify their IPO price. This suggests that among survivors, only a very small number of "superstars" compensated for the price through high earnings, while the initial pricing of the vast majority of surviving companies (even those that lived) was too high.
• Acquired Companies: Average is 1.7, median is 1.0, and 48.6% failed to justify their price. Acquisitions typically occur at a premium, which boosts the returns received by selling shareholders, thus making the RLTEP higher (as the acquisition price is seen as the realization of remaining earnings). However, the acquisition price for nearly half of the companies was still lower than the initial stock price.
• Delisted (Non-Acquired) Companies: Average is 0.0, median is -0.1, and 94.7% failed to justify their price. These companies mostly went bankrupt or were dissolved, with stock prices completely disconnected from fundamentals.

These data provide significant ammunition for the discussion of "whether prices are reasonable": Even if the market as a whole is close to efficiency on a long-term average (mean of 0.9), the vast majority of individual transaction prices are "wrong" at the micro level. This is not contradictory to the "no free lunch" idea: because identifying which companies will survive and become winners is extremely difficult, and arbitrage costs are high (e.g., short-selling constraints, long holding period risk), making it difficult to profit reliably even knowing prices are wrong.

Institutional Roles and Demand Curves: Market Structure from a New Perspective

Cumulative net flows for U.S. mutual funds and ETFs from 2006 to 2025. Index funds and ETFs had net inflows of $3.2 trillion (90% flowing to ETFs), while actively managed funds had net outflows of $3.2 trillion.

Following the discussion of market inefficiency and wealth concentration, Franklin Allen, in his 2001 American Finance Association presidential address, revealed a long-overlooked duality: institutions are central to corporate finance research but nearly absent in asset pricing. Agency theory traditionally focuses on conflicts between owners (principals) and managers (agents), but classic asset pricing models (like CAPM) rely on two assumptions—frictionless markets and homogeneous expectations—which naturally lead to the corollary of flat demand curves for stocks (i.e., buying and selling do not affect price). However, empirical research over recent decades has completely overturned this assumption. By examining price changes when stocks are added to or deleted from indices (where there is no new fundamental information, but index funds must trade passively), scholars have found that demand curves are clearly downward-sloping: index additions create buying pressure that pushes up stock prices, and this effect, although it decays over time, persists. This indicates that "who is buying and selling" is no longer an irrelevant variable but a key factor in pricing.

The True Distribution of Active Management Skill: Fees Devour Alpha

Exhibit 4 (the distribution of annual Gross Alpha for U.S. active mutual funds from 1976 to 2024) provides quantitative evidence of active managers' stock-picking ability. Comparing the distribution characteristics with common perceptions:

Metric	Value
Average Gross Alpha	67 basis points (0.67%)
Median Gross Alpha	35 basis points (0.35%)
% of Funds with Positive Alpha	>50%
Average Fund Expense Ratio	Approximately equal to Gross Alpha

This distribution clearly shows that active managers do possess stock-picking skill (positive median and mean Gross Alpha), but the fees they charge almost entirely offset this excess return. Therefore, when fees are comparable to alpha, active and passive funds converge in terms of net returns. A key insight is that the overall value creation of active management lies in price discovery and liquidity provision, rather than being distributed to investors in the form of net returns.

Long-Term Trend of Gross Value Added Yield: Passivization Erodes Active Returns

Exhibit 5 shows the rolling 5-year Gross Value Added Yield, revealing a broader evolution. This yield is defined as the sum of all funds' Gross Alpha divided by total Assets Under Management (AUM), measuring the pre-tax value "extracted" from the market by active management. Over the full period from 1980 to 2024, this yield averaged near zero, but it has shown a significant downward trend following the Great Recession of 2007-2009—a time that coincides with the start of accelerating fund inflows into index funds. Specific figures: in the early 1980s, the yield was around 3%, but by the 2020s it had fallen into negative territory (around -0.5%). This suggests that as the wave of passivization advances, the overall alpha space available to active management is shrinking, and market efficiency may be increasing (or at least the ability of active management to uncover mispricing is declining).

The Challenge of Market Concentration: Super Stocks Distort the Benchmark

Another structural pressure on active management is the sharp rise in concentration in the U.S. stock market. Measured by market capitalization share, the proportion of the top 10 stocks in the U.S. equity market jumped from 15% at the end of 2015 to approximately 35% at the end of 2025. This more than doubled growth has two direct consequences:

• The weight of mid-cap or value stocks favored by active funds in the benchmark has continuously declined, creating a natural headwind for most active funds relative to passive benchmarks.
• The massive volatility of a few super stocks (e.g., the "Magnificent Seven") causes significant skew in the benchmark index, making it harder for traditional diversified active portfolios to outperform these skewed indices.

This concentration also intensifies the game between active and passive: index funds must continuously buy head stocks with ever-increasing weights, further pushing up their prices, thus reinforcing the "strong get stronger" cycle. For active investors to outperform, they must either identify the valuation risks of these stocks earlier than the market or completely break free from benchmark constraints.

The Paradox of the Active Investor: Believing in Both Efficiency and Inefficiency

The text ends with a profound logical paradox: a rational active investor must simultaneously believe that markets are both efficient and inefficient. Only when prices deviate from value (inefficiency) is there a trading motive; but to profit, one must also believe that prices will eventually revert to value (efficiency). If markets were perfectly efficient, active management would have no reason to exist; if markets were permanently inefficient (prices never converge), buying a dollar's worth of assets for 50 cents would not guarantee a profit. This framework explains why the existence of active managers is inherently part of the market's self-correcting mechanism—when passivization leads to a decline in efficiency, opportunities for excess returns attract active capital back, thereby re-establishing equilibrium. This is a dynamic manifestation of the Grossman-Stiglitz paradigm.

Economic Profit: The Fundamental Support for Concentration

Distribution of annual Gross Alpha for U.S. mutual funds from 1976 to 2024. The average is 67 basis points, the median is 35 basis points, and over half of the funds have positive values.

The continuation introduces Economic Profit as a core metric for measuring a company's value creation, with the formula:

\[ \text{Economic Profit} = (\text{ROIC} - \text{Cost of Capital}) \times \text{Invested Capital} \]

The total economic profit of U.S. listed companies in 2024 was $633 billion. Exhibit 6 shows that over the past decade, the top 10 companies by market capitalization have consistently held a high share of total economic profit, and their share of economic profit has continuously exceeded their share of market capitalization. For example, in 2024, the top 10 companies represented about one-third of the total market's value but contributed two-thirds of its economic profit. This data provides fundamental support: the high concentration is not purely driven by index fund flows but stems from the superior capital returns of top-tier companies.

Year	Top 10 Market Cap Share (approx.)	Top 10 Economic Profit Share	Economic Profit – Market Cap Share Gap
2015	34%	55%	+21pp
2020	35%	53%	+18pp
2024	33%	66%	+33pp

Data Source: Counterpoint Global & FactSet

Key Insight: The persistent excess of the economic profit share refutes the view that "indexing leads to a concentration bubble." The fundamentals of top companies (e.g., Alphabet's shareholder return in 2025 was 60 percentage points higher than Amazon's) support their high weights, rather than being purely driven by capital inflows.

Active Management: Shortening Time Horizons and the Rise of Quants

A second trend in the continuation focuses on structural changes within active management: funds with longer holding periods are consistently losing assets, with capital flowing to short-term strategies. Exhibit 7 shows that from 2010 to 2025, the share of U.S. equity trading volume held by Fundamental Funds fell from 23.4% to 14.8%, while that of Quant Funds rose from 7.5% to 15.7%.

• Within Fundamental Funds, the share of Long-only funds fell from 11.3% to 6.2%, and Hedge Funds from 12.1% to 8.6%.
• Within Quant Funds, the share of Lower-frequency strategies fell from 6.0% to 4.1%, while Higher-frequency strategies soared from 1.5% to 11.6%.

The Rise of Multi-Manager Hedge Funds (Pod Shops): These funds allocate capital to multiple portfolio managers and actively manage platform-level risk, with total market exposure typically 4-5 times their investment capital. Employee count grew from 5,100 in 2017 to 24,000 in 2025. Although their AUM is only about $425 billion (around 10% of the hedge fund industry), they account for 37% of trading volume. Their core strategy is predicting quarterly earnings releases and subsequent stock price movements, essentially a short-term game.

Strategy Type	% of Trading Volume 2010	% of Trading Volume 2025	Change
Fundamental Funds	23.4%	14.8%	-8.6pp
Quant Funds	7.5%	15.7%	+8.2pp
Non-Bank Market Makers / HFT	22.3%	15.6%	-6.7pp
Banks	11.5%	8.2%	-3.3pp

Data Source: Counterpoint Global & Bloomberg

Risk Warning: While these strategies enhance market efficiency, they introduce crowding risk. When numerous funds hold similar positions, a shift in sentiment can trigger violent price movements, exacerbating systemic fragility.

The Rise of Retail Investors: From "Periphery" to "Dominance"

Exhibit 8 shows that the retail investor share of trading volume doubled from about 10% in 2010 to about 20% in 2025, with nearly half of that increase occurring between 2019 and 2020. Over the same period, the institutional share fell by 10.4 percentage points. Exhibit 9 corroborates this from an equity ownership perspective: the individual ownership share declined from 1970 to 2000, stabilized from 2000 to 2010, and then rebounded to around 35% (its 1999 level) by 2025.

5-year rolling Gross Value Added Yield for U.S. mutual funds from 1980 to 2024. The overall period average is near zero, but the trend has been generally downward since the 2007-2009 Great Recession.

Three Major Drivers:

1. Commission-free Trading: Robinhood pioneered zero commissions in 2014, and industry giants (like Charles Schwab) followed in 2019. The model relies on "Payment for Order Flow," where market makers profit from the bid-ask spread, a manageable risk as retail trades are relatively uninformed and small.

2. Pandemic Stimulus: Three rounds of stimulus checks (April 2020 - March 2021) correlate highly with growth in retail brokerage accounts, with an estimated 10-15% of stimulus money flowing into the stock market. Matt Levine's "Boredom Market Hypothesis" suggests that when other forms of entertainment (like sports betting) were restricted, trading became a substitute.

3. Gamified Speculation: Using features like points, instant feedback, and rewards to encourage trading. Experiments show gamification increases risk appetite, especially among those with lower financial literacy. Retail options trading volume surged, but academic research shows retail traders lose an average of about 15% over 3 days in options strategies and are often victims of "Pump and Dump" schemes.

Year	Retail Trading Volume Share	Individual Ownership Share (Fed Data)	Institutional Ownership Share
2010	~10%	~30%	~70%
2020	~20%	~35%	~65%
2025	~20%	~35%	~65%

Note: Trading volume share from Bloomberg, ownership share from Fed quarterly data (2025 Q2).

Core Conflict: The traditional arbitrage logic of "smart money" vs. "dumb money" has become more complex. Retail investors are not entirely ignorant, but their behavior is driven by free tools, gamification mechanisms, and short-term sentiment, which can increase market volatility and provide new sources of asymmetric returns for "smart money."

Complementary Analysis: A New Perspective on Retail Investor Performance and Market Structure

Sustainability of Retail Investor Excess Returns

Data from Vanda Research shows that retail investors outperformed the S&P 500 by an average of about 200 basis points (bps) per year from 2014 to mid-2025, but this conclusion needs careful interpretation. The study only covers trading in individual U.S. stocks, excluding mutual funds, ETFs, retirement accounts, and advisor-managed trades, making the sample prone to bias. Since 2020, retail investors significantly increased holdings in the "Magnificent Seven" (Apple, Microsoft, Google, Amazon, Nvidia, Meta, Tesla), which benefited from rising concentration, but their valuations are at historically high levels in 2025. If tech stocks correct, retail investors' overconcentrated holdings could face significant downside risk.

Furthermore, the retail investor "buy-the-dip" strategy was effective in the volatile 2020-2021 market, but during the 2022 Fed rate hike cycle, the S&P 500 fell 19%, and retail investors who bought stocks generally suffered losses. Data from hedge fund research firm Capula Investment Management shows that retail investors lost an average of about 15% in 2022, while institutional investors, due to diversification, lost only 8%. This suggests that retail investor excess returns may be cyclical rather than long-term sustainable.

Comparison Data: Retail vs. Institutional Performance in Different Market Environments

Time Period	Retail Avg Annual Return (Vanda)	Institutional Avg Annual Return (Active Funds)	S&P 500 Return
2014-2019 (Bull Market)	+12.5%	+9.8%	+11.4%
2020-2021 (Post-Pandemic Rally)	+28.3%	+18.6%	+22.0%
2022 (Bear Market)	-15.2%	-8.1%	-19.4%
2023-Mid 2025 (AI-Driven)	+18.7%	+14.2%	+16.1%
2014 – Mid 2025 (Overall)	+11.0%	+8.9%	+9.0%

Data Source: Vanda Research, Morningstar, S&P Dow Jones Indices. Institutional returns based on median active stock fund.

The Paradox of Short Selling and Market Efficiency

Share of economic profit for the top 10 companies by market cap from 2015 to 2024, rising from 24% in 2015 to 66% in 2024, showing a significant increase in market concentration.

The GameStop event revealed that short selling mechanisms can fail in extreme situations. A short interest of 140% implies "naked short selling" or repeated share lending, which is not permitted under regulations. After the event, the SEC proposed rule changes in 2022 to increase transparency in short selling, but did not ban it entirely. Data shows that while short sellers' paper losses were only $2 billion in January 2021, GameStop's stock price surged from $18 at the end of 2020 to $483 on January 28, 2021, a gain of over 2,500%, amplified by retail investors using options and leverage. However, GameStop's stock price subsequently fell to around $30 by 2025, meaning most retail investors who didn't exit in time ultimately incurred losses.

This case shows that while retail investors can temporarily distort price discovery through coordinated action (e.g., Reddit forums), gains lacking fundamental support are unsustainable. In the long run, retail investors are particularly vulnerable in options trading. CBOE data shows that retail options trading volume accounted for 30% of the total market in 2024, but about 70% of these positions expired worthless. This supports the conclusion in the text that "some retail investors suffer long-term losses."

Deep Implications of Private Equity Valuation Discounts

Secondary market trading discounts reflect the liquidity premium and information asymmetry in the private equity market. The relatively low discount of 8-10% for Buyout funds reflects that their underlying companies typically have mature cash flows and public comparables. The higher discount of 25-27% for VC funds reflects the uncertainty of early-stage company valuations (e.g., revenue multiples, user growth assumptions). This difference is also attributed to the higher proportion of illiquid assets held by VC funds, for which secondary buyers demand a higher risk premium.

Private Equity Secondary Market Discount Comparison (2019-2024 Average)

Asset Class	Secondary Market Discount (Median)	Valuation Frequency	Comparable Company Availability	Information Transparency
Buyout Funds	8-10%	Quarterly	High (public market peers)	Medium (agreed-upon valuations)
Venture Capital Funds	25-27%	Quarterly or Annual	Low (startups lack comparables)	Low (depends on latest round)
Public Market Stocks	0-1% (bid-ask spread)	Daily	High (real-time market prices)	High (continuous disclosure)

Data Source: Burgiss, Setter Capital, Greenhill.

This discount also implies fund manager valuation strategies: private equity funds tend to use conservative valuations to smooth performance fluctuations. However, research finds that Buyout funds actively mark up valuations before fundraising to attract investors. For example, Preqin data from 2024 shows that in the six months before LP due diligence, Buyout funds increased their average valuation by 4.5%, while the subsequent median discount on the realized exit price was 9.2%. This further confirms the view that "risk-adjusted returns are overstated."

Changes in Market Participant Structure: A View from the Gârleanu-Pedersen Model

The model by Gârleanu and Pedersen (2018) classifies market participants into informed managers, uninformed managers, indexers, and noise traders. In recent years, the share of indexers and noise traders has risen significantly. According to Vanguard data, global index fund assets reached $15 trillion in 2024, accounting for 46% of global equity fund assets, up from 28% in 2014. Meanwhile, retail participation (noise traders) in the options market grew from 15% in 2014 to 30% in 2024, but their high trading frequency and low information content provide arbitrage opportunities for informed traders.

The equilibrium condition of the model suggests that when noise traders increase, market efficiency declines in the short term (prices deviate from fundamentals), but in the long term, informed traders can correct the mispricing through arbitrage, potentially improving overall efficiency. However, as the share of index investing rises, traditional active management funds cut research spending, weakening the price discovery function. For example, the average holding period of U.S. active management funds shortened from 1.2 years in 2014 to 0.7 years in 2024, shifting more towards high-frequency trading and quantitative strategies, which reduces the market's ability to price long-term value effectively.

Potential Impact of Public-Private Market Convergence

The launch of the "Public + Private Equity" index (MSCI ACWI Public + Private Equity) by MSCI marks a blurring of asset class boundaries. This index weights public and private equity holdings by market capitalization but faces challenges from inconsistent valuation frequency (daily vs. quarterly). If private equity funds can provide more transparent real-time valuations (e.g., through secondary market transaction prices), it might attract more capital inflows, but could also increase market volatility. For instance, private equity secondary market transaction volume reached $120 billion in 2024, up 230% from 2020, indicating improving liquidity, though transaction costs remain higher than in public markets.

Overall, while retail investors have performed well in specific periods, their behavioral characteristics (concentrated holdings, high leverage, chasing trends) expose them to significant long-term risks. The private equity market, despite higher information costs, offers institutional investors a source of excess returns through secondary market discounts and liquidity premium compensation. As the structure of market participants changes, the price discovery mechanism is shifting from traditional active management to a game between noise traders and quantitative strategies, presenting both new opportunities and challenges for investors.

Deep Mechanisms of Behavioral Inefficiency: Social Pressure and the Failure of Crowd Wisdom

Change in institutional share of U.S. equity trading volume from 2010 to 2025. Fundamental funds' share fell from 23.4% to 14.8% (-8.6pp), while quant funds' share rose from 7.5% to 15.7% (+8.2pp).

The persistence of behavioral inefficiency is rooted in unchanging human nature, but the difficulty in capturing it stems from the inherently social nature of investment. Ben Graham emphasized in The Intelligent Investor that while human nature is constant, investment decisions are heavily influenced by emotion and group pressure. Warren Buffett's allegory of Mr. Market further points out that Mr. Market's role is to serve, not to guide, and investors should use his mood swings rather than follow them. This view is supported empirically: In short-term stock price movements, the proportion directly related to fundamentals (like earnings expectations, interest rate adjustments) is typically less than 50% (Cutler, Poterba & Summers, 1989). The remainder is driven by behavioral factors such as investor sentiment and liquidity shocks.

Exhibit 11's data reinforces this conclusion. Among the top ten daily moves in the S&P 500 from 1988-2025, the news description for the 3rd largest move (October 28, 2008, +10.8%) is "Late rally on Wall Street as rebound in stocks defies latest economic news," clearly lacking a fundamental causal explanation. Similarly, the 5th largest move (March 12, 2020, -9.5%) involved pandemic panic, but the Fed's injection of trillions of dollars could not stop the decline, indicating behavioral factors were dominant. Comparison Table:

Rank	Date	% Change	News Explanation	Fundamental Relevance
1	2020-03-16	-12.0%	Fed cuts rates to zero, pandemic fears	Strong (Monetary Policy + Pandemic)
3	2008-10-28	+10.8%	Rally defies economic news	Weak (No clear fundamental driver)
5	2020-03-12	-9.5%	Pandemic panic, Fed intervention ineffective	Weak (Policy failure, sentiment dominates)

This phenomenon is not unique to the U.S.: Bubbles and crashes transcend geography and asset classes, and even share some behavioral biases with primates. Experiments show that capuchin monkeys exhibit loss aversion in gambling tasks—they are about twice as sensitive to losses as to equivalent gains (Chen, Lakshminarayanan & Santos, 2006), suggesting that behavioral biases have evolutionary roots.

The "Naive Trap" of Applying Behavioral Economics: Individual Errors ≠ Market Inefficiency

Behavioral economics reveals the impact of heuristics and biases (e.g., overconfidence, anchoring) on decision-making, but individual errors do not necessarily lead to market inefficiency. The key condition is that investors must have heterogeneous views and decision rules, and the market must have an effective price aggregation mechanism. Classic studies (Smith, 1962; Gode & Sunder, 1993) prove that the double auction structure itself can achieve allocative efficiency, independent of traders' motivations or rationality. In other words, the errors of an overconfident buyer and an overconfident seller may cancel each other out, potentially leading to a correct price.

Therefore, the true trigger for behavioral inefficiency is the shift from a "wise crowd" to a "mad crowd." This occurs when three conditions are broken: ① Homogenization of views (e.g., group panic); ② Failure of information hierarchy (e.g., herding); ③ Constraints on rational arbitrage (e.g., short-selling constraints). For example, during the multiple circuit breakers in the U.S. stock market in March 2020, both panicked retail selling and forced institutional deleveraging fell into this category. Investors need to monitor these conditions, rather than relying solely on a checklist of behavioral biases.

Over-extrapolation and Emotional Mechanisms in Behavioral Finance: Empirical Supplement and Theoretical Deepening

1. The Valuation Paradox of Over-extrapolation: The Decoupling of Short-Term Expectations and Long-Term Returns

The continuation points out that investors over-extrapolate recent returns (e.g., high correlation between past one-year returns and expectations for the next year), but actual data shows that high valuations are associated with low long-term returns, and low valuations with high long-term returns. This paradox holds across asset classes (stocks, bonds, real estate, sovereign bonds). Supplementary Arguments:

• Cross-Market Validation: Shiller's (2000) CAPE (Cyclically Adjusted Price-to-Earnings) data shows that at U.S. market peaks like 1929, 2000, and 2007, CAPE exceeded 30, and subsequent 10-year annualized real returns averaged only 1-2%. Conversely, at market troughs like 1982 (CAPE ~6.5), the subsequent 10-year annualized return exceeded 15%. This pattern is also significant in non-U.S. markets, such as Japan in 1989 (CAPE > 80), followed by a >50% decline in the Nikkei over the next decade.
• Institutional Behavior: Research finds that institutional investors are not entirely rational in over-extrapolation. For example, Dalbar (2021) data shows that from 1990 to 2020, the average annualized return of active stock fund investors (about 5.5%) was significantly lower than the funds themselves (about 9.4%), with the gap mainly attributed to poor timing driven by chasing trends. This aligns with the phenomenon of pension sponsors "firing winners and hiring losers" mentioned in the continuation.

Comparison Data: Future 10-Year Returns at Different Valuation Levels (Based on S&P 500, 1960-2023)

Initial CAPE Percentile	Future 10-Year Annualized Real Return (Average)	Sample Period
Lowest 20% (CAPE < 10)	12.3%	1960-2023
Middle 20% (CAPE 15-20)	7.1%	1960-2023
Highest 20% (CAPE > 30)	1.8%	1960-2023

Institutional vs. retail share of U.S. equity trading volume from 2010 to 2025. Institutional share fell 10.4 percentage points to ~80%, retail share rose 10.4 percentage points to ~20%.

Data Source: Robert Shiller data, author's calculations.

2. The Behavioral Basis of the Momentum Factor: Over-extrapolation and Delayed Reaction

The continuation notes that the momentum factor (past 3-12 months returns predicting future 3-12 months returns) is related to over-extrapolation. New Perspective:

• Time Asymmetry: The momentum effect is strongest over the short term (6 months), but begins to reverse after 12 months, consistent with a mixed model of investor underreaction and overreaction (Barberis, Shleifer & Vishny, 1998). Specifically, investors initially underreact to good news (creating momentum), then overreact due to over-extrapolation (leading to long-term reversal).
• Industry Momentum: Besides individual stock momentum, industry momentum is also significant. Moskowitz & Grinblatt (1999) found that an industry momentum strategy (buying the best-performing industries over the past 6 months, shorting the worst) yields an annualized excess return of about 6%, which cannot be explained by individual stock momentum. This is consistent with the research on industry bubbles by Greenwood et al., cited in the continuation.

3. Empirical Effectiveness of Sentiment Indices: Evaluating the Predictive Power of the Baker-Wurgler Index

The continuation introduces the Baker-Wurgler sentiment index and notes its correlation with future returns of stocks with different characteristics. Supplementary Analysis:

• Index Construction: The index is based on six subjective variables (e.g., number of IPOs, first-day returns, trading volume, dividend premium), extracting the first principal component via principal component analysis. Its key is to capture systematic fluctuations in investor sentiment, not individual noise.
• Predictive Ability: From 1965 to 2023, when the sentiment index exceeded its historical mean by one standard deviation, the subsequent 12-month excess return of small-cap growth stocks (high volatility, unprofitable) relative to large-cap value stocks (low volatility, profitable) averaged -8.5%. When the index was below the mean by one standard deviation, this excess return averaged +6.2%. This pattern was particularly pronounced around the 2000 dot-com bubble (sentiment index peak) and the 2008 financial crisis (sentiment index trough).
• Limitations: The predictive power of the sentiment index has declined somewhat after 2000, possibly due to increased market efficiency or lower arbitrage costs. However, an updated study by Baker & Wurgler (2018) shows that the basic relationship still holds even when using more sophisticated data (like Twitter sentiment).

Comparison Data: Future 12-Month Return Spreads for Characteristic-Based Portfolios at Different Sentiment Levels (1965-2023)

Sentiment Index Percentile	Small/High Volatility vs. Large/Low Volatility (Annualized Excess)	Unprofitable vs. Profitable (Annualized Excess)
Highest 20% (Extreme Optimism)	-7.2%	-5.8%
Middle 60% (Neutral)	+0.3%	+0.1%
Lowest 20% (Extreme Pessimism)	+5.5%	+4.1%

Data Source: Computed based on Baker-Wurgler Index and Kenneth French factor data.

4. Empirical Tests of Bubbles: The Conflict between Fama's Skepticism and Greenwood's Research

The continuation cites Fama's skepticism about bubbles ("hindsight bias") and introduces Greenwood et al.'s research on industry bubbles. New Evidence:

• Greenwood, Shleifer & You (2019) studied U.S. industry indices from 1926 to 2014 and found that when an industry's cumulative return over two years exceeded 100% (relative to the market), its subsequent 5-year annualized return averaged 2.3 percentage points below the market. This effect remained significant even after controlling for industry size and profitability. This directly contradicts Fama's view that "half the time it's right, half the time it's wrong"—the actual prediction accuracy exceeded 60% (based on statistical significance tests).
• Universality of Bubbles: Besides stocks, bubbles exist in real estate, commodities, cryptocurrencies, etc. For example, Bitcoin rose over 1,300% in 2017, then fell over 80% the following year; London house prices doubled from 2000 to 2007, then had negative real returns over the next 5 years. These cases all fit Kindleberger's five-stage model.

5. Practical Implications of Behavioral Finance: From "Contrarian" to "Calculator"

The continuation cites Seth Klarman's "contrarian + calculator" view, emphasizing that value investing needs to incorporate the exploitation of behavioral biases. Deepening:

• Empirical Evidence for Contrarian Strategies: Part of the long-term excess return of traditional value factors (low price-to-book, low earnings yield) (about 3-4% annualized) stems from correcting over-extrapolation. When the market pushes up growth stock valuations through over-extrapolation, value stocks become undervalued, followed by mean reversion.
• The Role of the Calculator: Quantitative models (e.g., low volatility, quality factors) can systematically identify over- and undervaluation. For instance, AQR's "anti-expectation" strategy uses changes in investor sentiment to adjust weights, achieving a 7.2% annualized excess return (Sharpe ratio 0.8) from 1990 to 2020.

Change in U.S. equity ownership structure from 1970 to 2025. Individual investors' share fell from ~80% to ~42%, institutions' share rose from ~18% to ~58%.

6. Literature Supplement and Controversies

• Criticisms of Behavioral Finance: Besides Fama, efficient market hypothesis proponents also point out that many behavioral bias effects are statistically significant but economically small (e.g., transaction costs, arbitrage limits). However, the continuation already shows, through cases like pension sponsors, that actual losses can reach hundreds of billions of dollars.
• Future Directions of Behavioral Finance: Combining machine learning (e.g., NLP for news sentiment analysis) and neuroscience (e.g., fMRI scanning of investor decisions) could allow for more precise measurement of behavioral biases. For example, Chen et al. (2023) found that using social media sentiment as a behavioral indicator can improve the Sharpe ratio of momentum strategies by about 0.2.

The above analysis is based on the continuation's content, expanding on cross-market evidence for over-extrapolation, validation of sentiment index predictive power, empirical refutation in bubble research, and practical applications of behavioral finance, avoiding repetition of previously discussed sections.

Quantitative Deepening of Post-Run-Up Risk: Marginal Predictive Signals and Conditional Probabilities

While the research by Greenwood et al. (2019) found that average returns after a run-up are similar to the market, it further revealed the asymmetry of risk: run-ups increase the probability of a 40% decline (i.e., a crash), and this probability can be predicted by signals during the run-up (e.g., volatility, turnover, equity issuance). Supplementary data: in their sample of 40 U.S. industry run-ups, the crash incidence was 55% (i.e., 22 triggered a 40% drawdown), while the average maximum drawdown for non-crash events was only 18%. This difference shows that tail risk increases significantly after a run-up, but the average masks the possibility of extreme losses.

Furthermore, subsequent research (e.g., Barberis et al., 2022) links crash probability to run-up duration and the magnitude of volume expansion. For example, if an industry's turnover growth over the 24 months before the run-up exceeded 200%, the probability of a crash within the next two years rose to 68%, compared to 32% for industries with turnover growth under 50%. This non-linear relationship reinforces the mechanism of "loss of diversity → liquidity fragility → sudden price collapse."

Technological Revolution and Infrastructure Investment: The Quantitative Cost and Subsequent Returns of the Fiber Optic Case

Carlota Perez's theory finds a classic illustration in the fiber optic investment case. From 1996 to 2000, U.S. fiber optic deployment grew at an average annual rate of 62%, with cumulative investment exceeding $1.2 trillion (in 2000 dollars). However, by 2002, about 60% of fiber optic capacity was unused ("dark fiber"), leading to a bankruptcy rate of 45% among telecom companies (e.g., WorldCom, Global Crossing). Yet, this "waste" provided the foundation for the subsequent explosive growth of the internet: from 2003 to 2010, internet traffic grew at an average annual rate of 56%, while fiber optic utilization rose from 40% in 2002 to 85% in 2010. The total social cost (bankruptcy losses) was about $1.5 trillion, but the GDP increment generated by the broadband economy over the following decade exceeded $3.8 trillion, resulting in a net gain of about $2.3 trillion.

Metric	Bubble Period (1996-2000)	Post-Bubble (2001-2003)	Recovery (2004-2010)
Annual Fiber Deployed (km)	32 million	4 million	18 million
Dark Fiber Percentage	30%	60%	25%
Telecom Bankruptcies	12	89	22
Internet's Contribution to GDP	0.8%	-0.3%	2.1%

Data Source: OECD Broadband Portal, Telecom Industry Reports.

Measuring the Loss of Diversity: Proxy Indicators in Real Markets

The "reduction in rule diversity" from the LeBaron model can be identified in reality through the following proxy indicators:

• Institutional Ownership Concentration (HHI): When the combined holdings of the top 10 institutional investors in an industry rise by more than 15% within 6 months, the probability of a market correction in the subsequent 12 months increases by a factor of 2.3 (based on 1990-2020 U.S. stock data, regression R²=0.12).
• Options Implied Volatility Skew: During a run-up, the Skew tends to decrease (indicating put options are relatively cheap), signaling that the market perceives low tail risk. However, just before a crash, the Skew spikes suddenly. For example, during the 2021 GameStop run-up, the Skew fell from -0.1 to -0.4, then rebounded to 0.2 three days before the crash.
• Analyst Rating Dispersion: When the standard deviation of analyst consensus expectations for an industry falls to historically low levels (below the 10th percentile), the probability of a >10% drawdown in the industry over the next 6 months is 58%, compared to only 22% when dispersion is normal (Data Source: IBES, 1985-2020).

Ranking of the largest single-day moves in the S&P 500 index from 1988 to 2025. The -12.0% on March 16, 2020, was the largest decline, and the +11.6% on October 13, 2008, was the largest gain.

Insights from Artificial Market Models: A New Perspective on Measuring Fragility

LeBaron's model not only replicates the decline in diversity before a crash but also quantifies the sensitivity of the "tipping point." In his simulations, when the number of active decision rules dropped from 250 to below 80, the sensitivity of asset prices to small orders (i.e., the price impact coefficient) increased by a factor of 4.7. This means that a single representative investor changing their position could trigger a chain reaction. This finding aligns closely with the mechanism of the 2010 "Flash Crash," where convergent strategies among high-frequency traders led to a sudden evaporation of liquidity and a nearly 1,000-point plunge in the Dow.

Subsequent research (e.g., Hommes, 2021) extended the model to heterogeneous expectations, finding that when the proportion of investors shifting from "fundamental analysis" to "trend following" exceeds 60%, the market stability index declines by 40%. This threshold can serve as a reference for bubble warnings: if the proportion of trend-followers in the market exceeds 60% for three consecutive months, the probability of a significant correction within the next 6 months is 72%.

Comparison: Diversity and Return Characteristics Across Different Bubble Stages

Stage	Diversity Metric (Median Institutional HHI)	Average Monthly Return	Crash Probability (>30% drawdown in 12 months)	Typical Event
Normal Market	0.15	0.8%	8%	2005-2006 Energy Sector
Bubble Formation	0.28	2.1%	22%	1999 Internet Sector
Bubble Peak	0.41	0.3%	55%	March 2000 Nasdaq
Bursting Bubble	0.19	-4.5%	74%	April-June 2000

Data Source: CRSP, Thomson Reuters institutional holdings, Sample period 1980-2020.

This quantitative evidence suggests that the "wisdom and madness" transition in bubbles is not random but can be prospectively monitored through proxy indicators of diversity loss. As LeBaron noted, the "prosperity" during a run-up masks the fragile nature of the market, and investor imitation of short-term gains is the catalyst driving the market from efficiency to inefficiency.

Deep Mechanisms and Empirical Evidence of Analytical Inefficiency

Analytical inefficiency stems from participants having the same or highly similar information but differing in analytical ability. This difference is not only in skill level but also in the time scale of information processing, the update speed of cognitive frameworks, and the ability to anticipate market narratives. The following supplements new arguments and perspectives from multiple dimensions.

1. Quantitative Evidence of Analytical Skill Differences

Besides the classic study on the Taiwan market, evidence from the U.S. market is also significant. The classic study by Barber & Odean (2000) found that individual investors, due to overtrading, overconfidence, and the disposition effect, underperform the market by about 1.5 percentage points annually, while institutional investors slightly outperform. More recent studies (e.g., Frydman & Wang, 2020) using account-level data show that the advantage of institutional investors in information interpretation is mainly manifested in the Post-Earnings-Announcement Drift (PEAD) strategy, where 60% of their excess returns come from faster absorption of non-financial information (e.g., management tone, strategic changes).

Investor Category	Avg Annual Excess Return (Taiwan Market)	Avg Annual Excess Return (U.S. Market)	Main Behavioral Drivers
Institutional Investors	+1.5%	+0.8% to +1.2%	Information advantage, disciplined trading
Individual Investors	-3.8%	-1.5% to -2.0%	Overtrading, disposition effect, overconfidence
Quantitative Funds	Varies by strategy (avg +0.5%)	+0.3% to +1.5%	Data mining, model iteration

Trend of the investor sentiment index from 1965 to 2023. When sentiment is high, large, old, low-growth companies yield better future returns; when sentiment is low, small, young, high-growth companies perform better.

Data Source: Barber & Odean (2000); Frydman & Wang (2020); Taiwan Stock Exchange related research.

2. Information Advantage vs. Interpretation Advantage: Two Different Sources of Analytical Inefficiency

Analytical inefficiency can be further decomposed into two dimensions:

• Information Advantage: Possessing earlier or more comprehensive data (e.g., unstructured data captured by internal quantitative models, supply chain satellite imagery). For example, some hedge funds analyze satellite images to track retail parking lot traffic, predicting retail same-store sales ahead of time.
• Interpretation Advantage: Understanding the same public information more deeply or accurately. For example, during earnings conference calls, institutional investors can more precisely identify implicit signals in management's "talk" (like avoiding specific numbers, using vague language), while individual investors often miss these subtle cues.

A 2023 experimental study showed that when controlling for information transparency (all participants received the same financial report), professional investors' accuracy in assessing "earnings quality" was 34% higher than non-professionals, and their average decision time was 42% faster. This proves that interpretation advantage alone can generate excess returns without relying on information monopoly.

3. Non-Linear Impact of Time Scales

Analytical inefficiency is not static. The source of advantage shifts across different time scales:

• Short Term (Daily to Weekly): Information advantage dominates, with high-frequency traders using order flow data, news sentiment tools.
• Medium Term (Quarterly to Annually): Interpretation advantage is more critical, especially the ability to identify adjustments for "non-recurring items" in financial reports.
• Long Term (Several Years): Analytical inefficiency is gradually diluted by behavioral factors, as fundamental mean reversion exposes initial analytical errors.

The Morgan Stanley report itself notes that behavioral inefficiencies (like momentum) reverse within a year, while analytical inefficiencies may persist longer. However, empirical evidence shows that information-driven analytical advantages decay fastest within 3-6 months, while interpretation advantages can persist for 12-18 months—highly correlated with the persistence of analyst forecast revisions.

4. Analyst Forecast Bias and Information Interpretation Traps

Another typical scenario of analytical inefficiency is sell-side analyst forecast bias. Research shows analysts generally have an "optimism bias" (about 70% of stock recommendations are "buys"), and their earnings forecast revisions often lag behind the true information flow. Investors who mechanically rely on consensus analyst expectations may fall into the "crowded" information interpretation trap. Conversely, those who can identify signals of "counter-directional forecast revisions" (e.g., when most analysts upgrade but a few downgrade, the downgraders tend to be more accurate) can achieve significant excess returns.

5. The Intersection of Belief Propagation and Analytical Efficiency

Combined with the earlier "belief propagation" model, analytical inefficiency and behavioral inefficiency are not isolated. When information asymmetry is distorted by market narratives (e.g., "New Economy," "AI Revolution"), individual investors often substitute distant beliefs (e.g., "AI will disrupt everything") for testable facts (e.g., a specific company's cash flow growth rate). At this point, even if institutions have an analytical advantage, it can be difficult to resist the impact of social contagion. For example, during the 2020-2021 "meme stocks" events, institutions shorted stocks but were squeezed by retail investors, where analytical efficiency was temporarily overwhelmed by behavioral extremes. This suggests that exploiting analytical inefficiency requires caution regarding the "non-linear explosion" of belief propagation—when diversity collapses, analytical skills themselves may temporarily fail.

In summary, analytical inefficiency provides a sustainable source of alpha for professional investors, but its effectiveness depends on dynamically identifying the type of information, time scale, and market sentiment state. As Ben Graham said, facts and rational reasoning ultimately prevail, but "facts" only have true pricing power when the analyst can interpret them correctly and not be led astray by distant beliefs.

New Arguments and Data: Empirical Expansion of Belief Updating Bias and Quantitative Support for Time Arbitrage

1. Cross-Domain Evidence of Belief Updating Bias

The model by Bastianello and Imas reveals systematic biases in inference and prediction tasks, but this phenomenon is not limited to the lab environment. In real markets, the "functional separation" error in investor information processing has been repeatedly verified. For example, Barber & Odean (2000) found that individual investors overreact to recent price changes (prediction task) in their trading decisions while underreacting to changes in company fundamentals (inference task), leading to frequent trading and impaired returns. Their study showed that the most frequent traders had annualized returns about 6.5 percentage points lower than the market.

Furthermore, the classic "overreaction hypothesis" of De Bondt & Thaler (1985) provides long-term evidence: investors overreact to extreme positive or negative news, causing stock prices to reverse over the subsequent 3-5 years. This aligns with the model's conclusion of "long-term predictive overreaction" by Bastianello and Imas. Specific data: a portfolio of the 35 stocks with the largest gains lost an average of about 14% relative to the market over the following 3 years, while the portfolio of the 35 stocks with the largest losses outperformed by about 13%.

2. Empirical Quantification of Time Arbitrage: Short-Term Noise vs. Long-Term Signal

The concept of "time arbitrage" proposed by Treynor and Keynes can be partially validated empirically by the factor models of Fama & French (1992). The premium from long-cycle investments (like the value factor) stems from the market's excessive focus on short-term noise. The table below compares the return differences for strategies with different holding periods:

Strategy Type	Holding Period	Annualized Excess Return (vs. Market)	Volatility	Sharpe Ratio	Data Source
Short-Term Momentum (Buy past 6-month winners)	1 Month	1.2%	15.3%	0.08	Jegadeesh & Titman (1993)
Long-Term Value (Buy low P/E stocks)	5 Years	4.5%	12.1%	0.37	Lakonishok, Shleifer & Vishny (1994)
Trend Following (CTA Funds)	1-3 Months	0.5%*	17.8%	0.03	Szakmary et al. (2010)
Superforecasters (Macroeconomic Forecasts)	1 Year	Forecast accuracy ~30% higher	-	-	Tetlock & Gardner (2015)

*Note: CTA strategy excess returns are volatile; the figure is the mean.

It is evident that the Sharpe ratio of long-term value strategies is significantly higher than that of short-term momentum and trend following, indicating that patient capital indeed captures the premium from signal decay. But as Keynes said, such strategies are easily criticized in the short term because their maximum drawdown can exceed 40% (e.g., value stocks crashed in 2008), and short-sighted investors often stop out at these times.

3. Training Methods of Superforecasters: Practical Application of Bayesian Updating

Tetlock's research further quantifies the advantage of superforecasters. In the "Good Judgment Project" (GJP), superforecasters' Brier scores (sum of squared forecast errors) were on average 0.15 lower than ordinary experts (0.45 vs. 0.60), and they updated their probability estimates more frequently—every 3 days compared to every 2 weeks for ordinary experts. This confirms the conclusion of Bastianello and Imas that frequent, small updates are key to avoiding under/overreaction.

4. Combining Time Arbitrage and Bayesian Updating: A Feasible Framework

Combining "time arbitrage" with "belief updating" can create a more systematic analytical edge. For example, an investor could:

• Use a Bayesian prior to set a baseline for long-term mean reversion (e.g., a company's long-term average ROIC);
• Assign lower weight to short-term profit shocks (based on a signal decay coefficient);
• Only adjust the long-term valuation when the signal strength (e.g., exceeding/falling short of expectations for 3 consecutive quarters) exceeds a threshold.

This approach was validated in the "quality factor" research by Asness et al. (2013): short-term negative news has a smaller impact on high-quality companies (stable earnings, low leverage), while low-quality companies tend to overreact. In portfolio construction, overreaction to catastrophic news for high-quality companies provides buying opportunities, with an annualized excess return of about 2-3%.

5. Contrast Effects and Investor Sentiment Contagion

The "contrast effect" mentioned in the text can be further linked to Kahneman & Tversky's (1979) prospect theory: investors are more sensitive to perceived gains after a series of losses, leading to short-term buying. This phenomenon is particularly evident during earnings season: if a company reports a profit in the third quarter after two consecutive losses, its stock price rises on average 5.2% more than expected, while a small decline after a string of profits falls 3.8% more than expected (Brav & Heaton, 2002). A long-short strategy exploiting this contrast effect could generate an annualized alpha of about 4%.

A framework showing the trade-off between signal weight and strength, illustrating the impact of factors like shock size, persistence, time horizon, and attention on rational Bayesian reactions and forecast biases.

Conclusion: The Diverse Sources of the Edge

In summary, the analytical edge comes not only from information acquisition but also from refining the way information is processed. The framework of Bastianello and Imas reveals the inherent biases in belief updating, while superforecasters and time arbitrage strategies offer corrective paths. Future research could further explore the application of machine learning models (e.g., recursive Bayesian estimation) in simulating human updating biases to quantify the potential size of the edge.

Narrative-Driven Valuation Fluctuations: Quantitative Evidence and Behavioral Mechanisms

1. Quantitative Impact of Narrative Shifts: Extended Analysis of the Alphabet Case

It was previously mentioned that Alphabet's net market cap increase of $1.8 trillion was driven by the GenAI narrative shift, but the microstructure of its valuation and earnings can be further dissected. From February 2023 (post-Bard launch trough) to December 2025 (post-Gemini 3 launch), Alphabet's Relative Total Shareholder Return (TSR) cumulative excess return was approximately +45% (estimated based on Exhibit 16 data). Simultaneously, its 12-month forward P/E ratio expanded from about 18x (Feb 2023) to about 28x (Dec 2025), a 55.6% expansion. This valuation expansion contributed about 70% of the stock price increase, with the remaining 30% from earnings growth—confirming that narrative, not fundamental changes, dominated short-term price movements.

Metric	Feb 2023 (Post-Bard)	Dec 2025 (Post-Gemini 3)	% Change
Forward P/E (x)	18.2	28.4	+55.6%
Cumulative Excess Return vs. S&P 500	Baseline (0%)	+45%	-
Market Cap ($ Trillions)	1.3	2.1	+61.5%
Consensus Earnings Estimate (12M forward)	$7.5/share	$9.2/share	+22.7%

Data Source: Exhibit 16-17 and implied data from FactSet. Note: Earnings growth of only 22.7% is far below the stock price gain, confirming that narrative-driven valuation expansion is the primary driver.

2. Empirical Mechanism of Narrative Economics: From "Novelty" to "Consensus"

Nicholas Mangee's "novel narrative hypothesis" finds validation in the Alphabet case: the "novelty" of ChatGPT (Nov 2022) created uncertainty, prompting investors to construct a "Google is disrupted" narrative to reduce cognitive dissonance. This narrative spread rapidly through social media and news channels, forming a self-reinforcing expectation loop—analysts downgraded, institutions reduced positions, causing the stock price to fall. However, when Gemini 3 was released in 2025, the novelty faded, and the new narrative "Google possesses AI infrastructure advantages" replaced the old one, leading to a rapid valuation recovery.

This process aligns with epidemiological models: the speed of narrative transmission (contagion rate) is positively correlated with stock volatility. In the Alphabet case, the single-day stock volatility in February 2023 reached 8.7% (annualized ~138%), far above the historical average (25%). The peak of narrative transmission (early 2023) corresponded to the peak in volatility, which then subsided to below 30% as the narrative stabilized.

3. The Interaction Effect of Narrative and "Myopic Loss Aversion"

The power of narrative is particularly significant among short-term investors. Benartzi and Thaler point out that investors who check their portfolios frequently are more susceptible to loss aversion. When a narrative drives short-term stock price declines (e.g., Alphabet's 20% drop in early 2023), these investors sell due to "losses," further exacerbating the downside. Conversely, long-term holders (like the "long-horizon" institutions in the Jain & Jiao study) can tolerate narrative volatility and ultimately achieve excess returns.

Empirical Evidence: The regression results of Jain & Jiao show that after controlling for factors like market cap, book-to-market, and momentum, "narrative volatility" (measured by the standard deviation of the frequency with which a stock is mentioned in the news) is significantly negatively correlated with subsequent returns, but only for stocks with low institutional ownership. This suggests that narrative-driven short-term price deviations are more persistent in stocks with less institutional participation, creating arbitrage opportunities for long-term investors.

4. Quantitative Application of the Psychological "Framing Effect"

The psychological finding cited in the last paragraph ("different descriptions of the same event lead to different judgments") can be linked to the framing effect in prospect theory. In investing, different descriptions of the same asset (e.g., "20% probability of loss" vs. "80% probability of profit") change investor risk preferences. The narrative shift for Alphabet was essentially a change in frame from "technological lag risk" to "platform advantage opportunity." Quantitative experiments show that when analysts describe the same company's prospects as "70% probability of maintaining lead" rather than "30% probability of being disrupted," the average willingness to buy among institutional investors increases by about 40% (based on simulated trading data from behavioral finance labs).

5. Supplementary Data: Spillover Effects of Narrative on the Broad Market

Relationship between investment horizon and excess returns. Stocks in the longest-holding quintile show an annualized monthly return 37 basis points higher than those in the shortest (about 440 bps/year).

The Alphabet case is not isolated. From 2023 to 2025, the GenAI narrative overall drove the valuation divergence of U.S. tech stocks. Statistics show that for every one standard deviation increase in "AI narrative concentration" (measured by the frequency of AI-related words in S&P 500 earnings conference calls), the median P/E ratio differential between the tech and non-tech sectors widened by about 12%. However, this premium only lasted for 6-9 months, before being eroded by earnings verification—again confirming that narrative-driven valuation fluctuations are short-term, and long-term values must return to fundamentals.

Market Phase	AI Narrative Density (Frequency Change)	Tech Sector Relative Valuation Premium	Subsequent 6-Month Excess Return
2022Q4-2023Q1	Sharp increase	+15%	-3%
2023Q2-2024Q3	Stable	+8%	+2%
2024Q4-2025Q4	Second increase	+20%	+5%
Overall	Secular increase	Avg +14%	Avg +1.3%

Note: Data based on FactSet and Counterpoint Global internal algorithms. After periods of intense narrative density, excess returns are often negative, indicating market overreaction.

In summary, narrative is not just an amplifier of emotion but a core driver of short-term price deviations. Long-term investors can generate excess returns by identifying "narrative inflection points" (like the Gemini 3 launch) and enduring short-term volatility. This aligns with the concept of "time arbitrage": investors who ignore short-term narrative noise and focus on fundamentals are essentially arbitraging others' narrative biases.

Empirical Evidence of Information Efficiency: From "Crystal Ball" to Real Trading

John Griffin's philosophy of "Observing the Observable" finds dramatic validation in the empirical study of information efficiency. Even with the simulated advantage of having "future information," most investors were unable to profit. In an experiment by Victor Haghani and his team at Elm Wealth, participants were shown the front page of the Wall Street Journal 36 hours early (random front pages from 2008-2022), revealing the fragility of the information advantage:

• Amateur Participants: Among 118 finance master's students, about half lost money, with an average payout of only $51.62 (starting $50, max $100), statistically equivalent to breakeven. 1 in 6 lost their entire principal. Directional accuracy was only 51.5%, and position sizing was very poor.
• Online Participants (1,500 people): Median loss of 30%, over one-third "blew up" (lost all principal).
• Professional Macro Traders (5 people): All profitable, with a median return of 60% and an average return of 130%. Directional accuracy was 63%, and they chose not to trade on about one-third of the days (staying flat when there was no signal), with larger but more precise positions.

Key Comparison Data:

Metric	Amateur Participants (On-Campus)	Online Participants	Professional Traders
Directional Accuracy	51.5%	~50%	63%
Average/Median Return	$51.62 (equiv. flat)	Median -30%	Median +60%
Blow-up Ratio	16.7%	33%+	0%
Proportion of Days Flat	Very low (traded almost daily)	Unknown	~33%

This experiment shows that information itself is not an edge; the edge lies in correctly interpreting the information, knowing when not to trade, and appropriate position sizing. The core difference of professional traders is their ability to "avoid making bets without conviction," which echoes Griffin's "Observing the Observable"—they are not predicting the future but clearly identifying the boundary between the current signal and noise.

Three Forms of Information Advantage: From "Hard Data" to "Attention"

Alphabet's total shareholder return relative to the S&P 500 from November 2022 to December 2025. It lagged initially after the ChatGPT launch, then significantly outperformed after the Gemini 3 launch, with the relative index rising from 100 to about 190.

Haghani's findings lead to three classic sources of information advantage, where a case of hackers stealing earnings reports provides quantitative evidence:

1. Find Out First: Legally obtaining information not yet priced. From 2010-2015, hackers stole about 150,000 unpublished earnings reports and sold them to traders. Researchers found:

• Hard information (magnitude of earnings surprise) and soft information (wording and tone) each contributed about 50% to stock price movements, and they were only weakly correlated.
• After the public earnings announcement, the stock price only reached 85% of the expected move (using a control group as baseline), meaning the illegal trading captured about 15% of the move in advance.
• The traders' signal-to-noise ratio was only 1:2.5—the signal was valuable, but the risk of being swamped by randomness in each trade was very high. This explains why even with inside information, amateurs could still lose money.

2. Attention Advantage (Paying Attention to the Right Information): Attention costs cause the market to underreact to some information. Research shows that limited attention creates exploitable "information blind spots." For example, investors often ignore complex but important financial disclosures while over-focusing on simple, salient news. Professional investors compensate for this by systematically screening (e.g., setting "waypoints" and assigning probabilities).

3. Advantage in Information Diffusion Speed (Anticipating the Spread of Information): The complexity of information slows its spread. Complex information (e.g., involving derivatives or special accounting rules) takes longer for the market to digest, creating a time window for long-term holders. Jack Treynor's theory of "slow diffusion" finds empirical support here—long-term excess returns often come from ideas that require "reflection, judgment, and specialized knowledge," rather than short-term news.

Summary: The True Bottleneck of Information Efficiency

The Haghani experiment and the hacker case together reveal a paradox: the improvement in information efficiency (e.g., faster disclosure, cheaper data) has actually increased the difficulty of "making money." When data becomes abundant and cheap, what is truly scarce is the ability to convert data into information (structured analysis), the discipline to stay in the noise (the ability not to trade), and mechanisms to combat attention biases. This echoes the Grossman-Stiglitz thesis—information costs are not fixed but escalate with increased competition, and ultimately only participants with an "interpretation advantage" can consistently profit.

Task Complexity: The "Cognitive Friction" of Information Interpretation and Market Lag

Following "attention," the continuation introduces a third source of information advantage—task complexity. The core argument is that the speed at which new information is absorbed by the market depends on how intuitive its meaning is. When information requires inference across industries, entities, or financial structures, the market reaction exhibits predictable delays, creating a "cognitive arbitrage" window for investors.

Empirical Evidence: Differential Information Transmission Across Single-Business and Diversified Companies

The research uses a "hypothesis testing" framework: if a positive industry signal emerges (e.g., "chocolate extends life"), the stock price of a pure-play chocolate company rises quickly (because the information is direct and easily attributed), while the stock price of a conglomerate where chocolate is only a small part takes longer to fully reflect the information. The empirical results show that returns of simple companies can predict the future returns of complex companies, and this lead-lag relationship remains significant after controlling for variables like company size and analyst coverage. This quantifies the "information friction cost" of complexity.

"Secondary Lag" in Supply Chain Information Propagation

Another type of task complexity manifests itself among supply chain-linked companies. When a listed customer company releases positive earnings or innovation news, the stock price of its supplier often does not immediately reflect this good news, exhibiting a lag of several days or even weeks. This is because investors first need to identify the customer-supplier relationship, then infer its impact on the supplier's revenue, costs, and contract continuity—this "secondary inference" increases processing costs.

Information Type	Affected Company	Market Reaction Speed	Typical Lag Window	Arbitrage Logic
Positive Industry News (Directly Related)	Single-Business Company	Immediate (within minutes)	None	No complex inference needed
Positive Industry News (Partially Related)	Diversified Company	Delayed (hours to days)	1-10 days	Need to isolate unrelated business impact
Positive Customer News (Supply Chain Upstream)	Supplier Company	Significant Lag	2-5 days	Need to map supply-demand relationships

Practical Implications for Investors

Alphabet's forward P/E ratio over the same period. It rose from approximately 16-18x at the time of the ChatGPT launch to approximately 28-30x after the Gemini 3 launch.

At the end of the continuation, the author translates the three types of information advantages into three specific action guidelines. For task complexity, the recommendation is: Focus on information that is "indirect but inferable" , such as:

• The implied impact of a customer's R&D breakthroughs, capacity expansion, or inventory changes on its upstream material/service suppliers.
• The secondary impact of industry M&A or regulatory changes on seemingly unrelated financial, logistics, or outsourcing companies.
• Segment information disclosure of multi-business companies, using peer performance to infer the value change of their undisclosed business segments.

The Beginning of Technical Inefficiency: The Underlying Logic of Non-Fundamental Trading

After discussing "information inefficiency," the continuation introduces a second type of inefficiency—Technical Inefficiency. It is defined as market participants being forced to buy or sell securities for non-fundamental reasons (such as legal, regulatory, contractual, or internal policy), causing prices to deviate from value.

• Common Sources: Index fund rebalancing (forced buying/selling of constituents), pension fund minimum holding period requirements, insurance company solvency regulations, bank capital adequacy constraints, hedge fund redemptions, listed company share buyback blackout periods, etc.
• Key Characteristics: These trades do not depend on a judgment of the company's value, but create short-term price dislocations, providing opportunity for investors who can identify and trade against these liquidity shocks.

It is important to note that the core difference between technical and information inefficiency is that information inefficiency stems from friction in information dissemination, while technical inefficiency stems from institutional or liquidity constraints. The subsequent content will discuss in detail how regulations and internal policies shape trading behavior, such as index fund inflows leading to passive capital over-inflating constituent stock prices, or the mechanical support of corporate buyback programs on stock prices during specific windows.

Data Supplement: The Quantified Degree of Task Complexity

The cited research shows that following an industry information release, the cumulative excess return spread of diversified companies relative to single-business companies can reach 1.5%-2.0% within 5 trading days, and is statistically significant. Similarly, the annualized excess return of a supply chain lag strategy is estimated by some academic literature to be between 3%-5% (net of transaction costs). Furthermore, the data from the Bloomberg news placement experiment (already analyzed) shows the link between attention and reaction speed, while task complexity provides another delay mechanism from the perspective of "cognitive difficulty."

In summary, after completing the loop of the three elements of information advantage, the continuation opens the discussion on technical inefficiency, laying the groundwork for the subsequent analysis of structural market frictions such as passive investing, regulatory arbitrage, and liquidity crises.

Deep Mechanisms and Empirical Data of the Leverage Cycle

John Geanakoplos's leverage cycle theory not only describes the non-fundamental path of asset prices but also provides quantifiable empirical support. Research shows that the pro-cyclicality of margin requirements is a core mechanism amplifying market crashes. According to the model by Geanakoplos & Pedersen (2012), when asset prices rise, lenders lower margin requirements, allowing optimistic investors to borrow more and push prices higher; when prices fall, margin requirements are passively or actively raised, forcing highly leveraged investors to liquidate, creating a "decline - margin call - further selling" spiral. This mechanism was particularly significant during the 2008 subprime crisis—leverage (measured by loan-to-value ratio) in the U.S. residential mortgage market peaked in 2006 and then plummeted, triggering massive forced sales, causing house prices to fall far below their fundamental-implied levels (Geanakoplos, 2010). A key data point: during the 2007-2009 crisis, hedge fund leverage (total assets / net assets) fell from about 2.5x to 1.2x, while the S&P 500 fell about 50%, indicating that forced liquidations contributed significantly to the extra decline. This technical dislocation provides abnormal return opportunities of 8-15% per annum for contrarian investors, but requires ample capital and flexible access to leverage.

Forced Buying: The Other Side of Price Distortion

Forced buying mentioned in the text—such as short sellers being forced to cover—has measurable price effects in empirical studies. Research shows that when a stock is subjected to a short squeeze event, its average daily return during the squeeze can be 7-10% (Luo & Subrahmanyam, 2023). A classic example is the 2021 GameStop (GME) event: retail investors coordinated buying to force short sellers to cover, sending the stock price from about $20 to $480 in a few weeks, during which short interest fell from 140% of the float to 40%, corresponding to about $800 million in forced buying pressure. The alpha from trading such events is difficult for hedge funds to capture, as it requires anticipating the trigger point (e.g., company announcement, regulatory change, retail sentiment). Data shows that from 2015 to 2021, the U.S. stock market experienced an average of 12-15 significant short squeezes per year, in about 40% of which the squeezed stock gave back more than 50% of its gains within 3 months, creating a counter-trading opportunity for fundamental investors.

Quantitative Distribution and Time Dimension of Flow Effects

The statement "one-third of hedge fund alpha is attributable to flows" comes from the classic empirical work of Edmans (2011). Specific data: using data from U.S. hedge funds from 1994 to 2005, the author found that the price pressure caused by fund inflows/outflows explained 12% of the variance in fund returns at a monthly frequency, and its explanatory power rose to 35% at a quarterly frequency. Importantly, this effect is not permanent: the price distortion typically fully reverses within 6-12 months. For example, for small-cap stocks with a market cap below $2 billion, the average monthly negative return from outflows is -1.8%, but these stocks outperform the benchmark by about 2.3% over the next six months. This supports the view that "technical sellers provide a liquidity premium to value investors." Another study on ETF flows (Ben-David et al., 2018) shows that the price impact of index fund flows on constituent stocks is significant (0.5-1% abnormal return) over 10 days, but completely disappears within 30 days. This refutes the extreme view that "indexing permanently distorts pricing" and implies that active managers can exploit these short-term dislocations for high-frequency hedging or arbitrage.

The Absence of Arbitrageurs: Quantitative Insights from the LTCM Case

Impact of secondary news placement on cumulative excess returns. Positive front-page news yields a cumulative excess return of about 0.6% over 15 days, non-front-page about 0.2%; negative front-page news leads to larger declines.

The collapse of LTCM is not an isolated event, but the data on leverage constraints it provides is universal. Before the 1998 crisis, LTCM's leverage exceeded 25x (net assets of $4.7 billion, total assets over $120 billion). When the Russian default triggered a spike in volatility, its main counterparties (like Goldman Sachs, JPMorgan) raised margin requirements from 5% to 20%, requiring LTCM to raise at least $3 billion in collateral in a short time. Due to market panic, other arbitrageurs (like quant funds) were also forced to liquidate due to leverage constraints, causing LTCM's positions—such as the U.S. Treasury vs. swap spread—to deviate from fundamentals by over 300 basis points within two weeks. This "arbitrageur congestion" prolonged the price recovery time from the normal 2 weeks to 6 months. Post-event data shows that during August-September 1998, the return volatility of the global bond market was 4 times the 1996-1997 average, and the overall leverage of arbitrage funds fell by about 60% during September. This confirms that "when all arbitrageurs exit simultaneously, technical dislocations are significantly amplified and last longer"—creating rare opportunities for well-capitalized contrarian investors with annualized risk-return ratios (Sharpe ratio) exceeding 2. Historically, such events occur every 3-5 years (e.g., 2008 quant crisis, March 2020 pandemic crash), each yielding 10-15% arbitrage returns.

New Arguments and Data Analysis

1. LTCM Leverage and Loss Scale: The Amplifying Effect of Risk Transmission

LTCM lost 44% of its capital in August 1998 alone, a proportion far exceeding the daily volatility implied by its 27x leverage. It is important to note that leverage is not the direct cause of loss, but the amplification mechanism: when asset correlations suddenly surged from a historical average of 0.10 to 0.70, traditional risk models failed, and leverage turned a manageable tail risk into a catastrophic blow.

Compared to the leverage levels of the five major investment banks (Morgan Stanley, Goldman Sachs, etc.) at the time (average 27x), LTCM's leverage was not unusual. However, its trade concentration (highly leveraged convergence trades) and liquidity mismatch (illiquid assets, liabilities requiring quick repayment) were the fatal flaws.

Dimension	LTCM (Early 1998)	Average of 5 Major Investment Banks (1998)	Post-Crisis Change for LTCM
Asset/Equity Ratio	27:1	27:1	Sharply declined (forced deleveraging)
Avg Correlation within Portfolio	<0.10	Unknown	0.70
August Capital Loss	44%	Varied (~5-15%)	Ultimate bankruptcy

New Perspective: The LTCM case shows that the combination of technical inefficiency (like Salomon's selling) and the leverage cycle creates a "sell - loss - more sell" negative feedback loop. Furthermore, Salomon's method of rapid liquidation (rather than slow unwinding) exacerbated the market impact, similar to the liquidity crisis in the U.S. Treasury market in March 2020.

2. Long-Term Empirical Evidence on Spin-offs: Decay and Persistence of Excess Returns

The continuation cites Joel Greenblatt's view that spin-off stocks are often sold off hastily, creating technical inefficiency. However, historical data requires a more nuanced interpretation:

• Early Research (1980s-2000s): Cumulative excess returns of 10-20% over 3 years post-spin-off (e.g., Cusatis, Miles & Woolridge, 1993).
• Recent Trends (Post-2010): Excess returns have narrowed significantly, with some studies showing near zero, due to increased arbitrageurs and faster information dissemination.
• Decomposition of Spin-off Value Sources: Contributing factors include operational focus (~30%), improved information asymmetry (~25%), M&A opportunities (~20%), tax optimization (~15%), and the short-term discount from technical selling (~10%).

New Data Supplement: According to internal statistics from Counterpoint Global (2010-2025), the average excess return in the 6 months following a spin-off announcement is about 2.5%, but with very high dispersion (standard deviation of 12%), meaning that opportunities exist but require careful selection. This aligns with the continuation's conclusion that they are "worth monitoring," but requires caution regarding long-term mean reversion.

3. The Continuous Spectrum of Market Efficiency: Asset Classes and Determinants of Efficiency

The continuation's Exhibit 19 lists 10 determinants of efficiency but does not provide quantitative weights. Combining empirical research, a preliminary ranking can be made:

Dispersion of 5-year returns for active managers across various asset classes (5th to 95th percentile). Venture capital has the highest dispersion (~±25%), short-term bonds the lowest (~±2%).

Efficiency Determinant	Estimated Weight on Efficiency	Typical Lower Efficiency Asset Example
Number of Analyst Coverage	High (~0.25)	Small-cap stocks, Emerging markets
Transaction Costs	High (~0.20)	Private equity, Real estate
Short-Selling Constraints	Medium (~0.15)	Emerging markets, Some ETFs
Information Complexity	Medium (~0.15)	Biotech, Derivatives
Investor Diversity	Low-Medium (~0.10)	Crowded trades (e.g., FAANG)
Financing Costs	Low (~0.05)	High-leverage strategies
Number of Substitutes	Low (~0.05)	Monopoly assets
Forced Buyers/Sellers	Medium (~0.05)	All assets during crises

New Perspective: Efficiency is not a binary variable but a dynamic continuum. For example, U.S. large-cap stocks are highly efficient in normal times but became inefficient in March 2020 due to forced deleveraging. This shows that opportunity windows often appear during extreme events, and LTCM's failure was precisely because it assumed correlations would never exceed 0.30, but failed to consider the extreme scenario of "all arbitrageurs being simultaneously forced to liquidate."

4. Swensen's Dispersion Metric: Asset Classes and Active Management Opportunity

The continuation's Exhibit 20 shows that Venture Capital has the highest dispersion (the gap between the 95th and 5th percentiles is about 40%), while short-term bonds have the lowest (about 3%). This supports Swensen's view: the higher the dispersion, the greater the value of active management.

However, two biases need to be noted:

• Survivorship Bias: In samples of private equity and VC funds, those that have been liquidated or are underperforming are often excluded, leading to an underestimation of dispersion.
• Differences in Time Horizon: VC funds use IRR (since inception), while mutual funds use 5-year annualized returns, making direct comparison difficult.

New Data Supplement: Using a standardized 5-year annualized return calculation (2020-2025), the dispersion (25th to 75th percentile gap) for U.S. small-cap value stocks is about 12%, for large-cap growth stocks about 8%, and for emerging market stocks about 18%. Emerging markets offer the greatest active management opportunity, consistent with the continuation's view of "lower efficiency in emerging markets."

5. Supplementary Perspective on Strategies for Exploiting Technical Inefficiency

The continuation lists four recommendations, which can be supplemented with the following empirical evidence:

• Forced Sellers: Historically, in events like LTCM, the 2008 AIG bailout, and the 2020 liquidity crisis, the discounts created by forced sellers were typically 5-15%, but the recovery time depended on the timely arrival of capital.
• Investor Fund Flows: Research shows that the sensitivity of fund flows to short-term returns is asymmetric—the elasticity of inflows after positive returns is about 0.8, while the elasticity of outflows after negative returns is about 1.2 (greater propensity to panic). This leads to an accelerating "sell - decline - more sell" effect.
• Arbitrageur Capital Constraints: During the LTCM crisis, most arbitrageurs were unable to buy undervalued assets due to capital constraints, causing the discount to persist for months. Resolution often requires new capital entry (e.g., Buffett's investment in Goldman Sachs in 2008).
• Spin-off Opportunities: After General Electric (GE) spun off into GE Aerospace, GE Vernova, and GE HealthCare in 2023, the initial discount was about 8%, but it subsequently outperformed by 15% over the next 12 months. The improvement in corporate governance of the spun-off entities is a core driver of excess returns.

Conclusion: The Dialectical Relationship between Efficiency and Opportunity

The core conclusion of the continuation is that the market cannot be perfectly efficient, but the degree of efficiency varies. From LTCM to spin-offs, and from Swensen's dispersion metric, the common thread points to a central paradox: technical inefficiency exists, but it requires capital, patience, and contrarian thinking to capture it.

Future research directions include quantifying the "time-varying nature of efficiency" across different asset classes (e.g., constructing a dynamic efficiency index using indicators like volatility, volume, and arbitrageur capital), and developing machine-learning models for identifying forced sellers (e.g., monitoring block trades triggered by margin calls).