[월가아재] 헤지펀드가 버블인 것을 알면서도 사는 이유

Finance / Markets Context

Market level / timing

• Nasdaq index is described as being near its previous high.
• Jackson Hole is referenced as a key upcoming market focus.
• A “largest downsizing in 10 years” narrative is centered on:
  • Semiconductors
  • Large tech stocks
  • July is mentioned.

Bubble debate

• The discussion frames whether the environment is an AI bubble:
  • Some argue it is in an early bubble stage.
  • Others argue it is not a bubble yet because performance remains strong.

---

Study Covered: Dot-Com Era Hedge Funds vs. Bubble

Central paper

• The paper examines whether sophisticated investors/hedge funds behaved as if they could profit from mispricing during the dot-com bubble aftermath (the subtitle references Dec / “Doc”).

Authors / source

• Marcus Brunnermeier (Princeton)
• Stefan Nagel (London Business School at the time)
• Published in the Journal of Finance (noted as a top finance journal) in 2004.

Data source and mechanics

• Uses U.S. filing data:
  • Form 13F (subtitles misread as “1-teen F”)
  • Hedge funds with >$100m assets report equity holdings quarterly.

---

Instruments / Tickers / Assets Explicitly Mentioned

Tick ers

• Cisco (used as an example internet/tech stock in the Nasdaq bubble period)

Indices / markets

• Nasdaq
• NYSE (mentioned to indicate patterns were less distinct outside the most expensive Nasdaq subgroup)

Sectors / themes

• Semiconductors
• Large tech
• AI theme
• “Overvalued tech” definition (important: based on price-to-sales ratio (PSR), not traditional industry classification)
• Internet stocks (dominant share of the “overvalued tech” group)

---

Key Numbers and Metrics

Dot-com bubble period timing

• End of March 2000 is referenced as a key point (Nasdaq peak context).
  • Overvalued tech stocks were 31% of hedge fund portfolios.
  • Overvalued tech stocks in the market were 21%.

Most aggressive divergence (earlier signal)

• September 1999 is cited as the point of the largest divergence.
  • Hedge funds increased allocation: 16% → 29%
  • Market weight moved: 14% → 17%
  • Net relative increase:
    • Hedge funds: +13%
    • Market: +3%

Performance/holding pattern for the high-PSR group

• The “top PSR” group (“Top 20% PSR”) quadrupled in two years.
• Hedge funds then gave back more than half of the gains by the end of that two-year window.

Timing of reversal / peak handling

• Individual stocks peaked at different times:
  • Some already in 1999
  • Others around March 2000
• A stylized interpretation is used:
  • Hedge funds held about ~double the shares in the quarter before a stock’s peak versus the quarter after.

---

Short-Selling Specificity and Limits of 13F

• Short positions are not directly visible in 13F snapshots because:
  • Derivatives/shorts are not fully captured.
• Therefore, the authors use inference from returns.
• Specialized short-selling funds are used as a sanity check:
  • They become visibly negative only after July 1999
  • AUM size is about 0.3% of the hedge fund industry
  • This implies short-bets were limited in scale and short-lived

---

Hypothetical Replication / “Luck vs Skill” Control

• After controlling with matched pairs, excess/alpha-like differences are reported:
  • 4.5% in the first quarter after disclosure
  • 2.7% in the second quarter
• Statistical significance:
  • Passed 10% and 5% significance levels respectively
• The effect fades toward ~0% in the third and fourth quarters
• Caveat:
  • Sample size is 12 quarters (~3 years), implying limited statistical power (the authors acknowledge this).

---

Hedge Fund Examples (Tiger / Soros and Others)

Tiger Fund (Julian Robertson)

• The Tiger Fund allegedly liquidated tech exposure:
  • By 1999, it “dropped to zero”.
• Jaguar Fund (Tiger) outflows:
  • Redemption pressure intensifies near the end of 1999.
• Fund liquidation announcement:
  • March 30, 2000
  • Nasdaq peak cited as March 10
  • Liquidation occurred about ~20 days after the Nasdaq peak.

Soros Quantum Fund

• Q3 1999:
  • Tech weighting tripled from <20% to ~60% in one jump.
• Capital inflows peaked during the same period.
• Theme: “opposing position cost”
  • Even Soros later faces heavy outflows during tech collapse; the narrative suggests Quantum is not “safe” either.

Other cited hedge fund example (subtitle-distorted)

• A Drew/Drummiller/Joseph Miller–style example appears (names are distorted by transcription errors).
• Core claims:
  • Alleged $600 million loss in spring 1999 after shorting due to an overheated market view.
  • Re-entry in March 2000 after the trend persisted.
  • Loss escalated to about $3 billion.
• Main point: shorting near/at the peak can generate massive losses if the trend continues.

---

Methodology / Framework Extracted From the Paper Description

1) Define “overvalued tech” mechanically

• Rank all Nasdaq stocks by price-to-sales ratio (PSR).
• Select the top 20% PSR names as “overvalued tech”.

2) Reconstruct hedge fund behavior using Form 13F

• Use end-of-quarter holding snapshots:
  • Entry/exit within the quarter is imperfect.
• Limitations:
  • Shorts and many derivatives are not captured in 13F.
  • Public release delay is about ~45 days after quarter end.

3) Track portfolio weights vs. market

• For each quarter:
  • Compare hedge fund weight in overvalued tech vs the market weight in the same group.
• Examine timing around:
  • The Nasdaq peak (March 2000)
  • Earlier divergence (September 1999)

4) Return-based inference about shorting

• Since shorts aren’t visible, infer effects via fund returns.
• Compare with specialized short funds to sanity-check timing/scale.

5) Individual-stock peak alignment (“event study” style)

• For stocks in the overvalued group:
  • Align by each stock’s own peak date.
• Compare hedge fund share holdings before vs after each stock’s peak.

6) Luck vs skill test using matched pairs

• For each held stock, create comparison pairs with similar:
  • market size
  • PSR category
  • returns over prior 6 months
• Build 125 comparison groups.
• Compute:
  • (fund stock return − matched pair return)
• Evaluate persistence of “excess” returns across quarters.

---

Explicit Recommendations, Cautions, and Takeaways

Main empirical takeaway

• Smart money (hedge funds) did not simply ignore the bubble.
• They:
  • Overweighted overvalued tech names more than the market.
  • Often reduced holdings before peaks.

But the “process” can still fail

• Shorting is difficult because short sellers are in the minority:
  • Timing risk: being correct but too early can still be punished.
• Redemption pressure can force liquidation even if the “fundamentals call” is right (Tiger example).

Risk-management framing

• Timing is critical for both:
  • short bets
  • long avoidance
• The emphasis is on:
  • having an exit plan
  • ensuring the evidence validity remains intact

Evidence framework (closing guidance style)

• If evidence for a trade is wrong:
  • Cut losses and exit (even while still down)
• If evidence remains valid:
  • You may average down or hold (context-dependent)
• Criticism:
  • Entering trades without a basis can lead to “getting lost” in subsequent buy/sell decisions as price moves.

---

Current-Market Bridge: AI / Tech “Smart Money” Basket

Goldman Sachs “hedge fund VIP basket”

• Goldman Sachs is referenced as having a “hedge fund VIP basket”:
  • 50 core stocks frequently held by hedge funds in top 10 positions.

Reported positioning claims

• Hedge funds’ divestment of AI/tech is said to have reached near all-time highs in May (as presented in the narrative).
• In Q2 disclosures, hedge funds reportedly have more AI exposure than mutual funds.
• The basket delivered worst relative performance to stocks in July, followed by a sharp contraction later (framed as severe crowding unwind).

Interpretation / caution from the narrative

• Goldman desk view: could be a “reset of trade overcrowding,” not necessarily loss of confidence.
• AI exposure decreased from the Q2 peak but remains above the long-term average.
• Host argues dynamics differ from 1999 because:
  • passive/index funds play a larger role
  • mechanical buying/selling from index flows affects prices
  • active funds, option-dealer liquidation, term rebalancing, and emergency AI funds can all influence price
• Data limitation caution:
  • 13F captures observable equity holdings but may miss derivatives/option-based exposure.
  • The subtitle claim: “no longer any exposure through options or derivatives visible here.”

---

Disclosures / Disclaimers

• No explicit “not financial advice” line appears in the provided subtitles.
• The discussion repeatedly frames itself as evidence-based, but no formal regulatory disclaimer is shown in the text provided.

---

Presenters / Sources Mentioned

Presenter / host

• “[월가아재]” (name implied by the video title; not otherwise explicitly identified in subtitles)

Academic paper authors

• Marcus Brunnermeier
• Stefan Nagel

Journal / source

• Journal of Finance (paper published/credited as 2004)

Hedge fund figures cited

• George Soros (Soros “Quantum Fund”)
• Julian Robertson (Tiger Fund; “Jaguar Fund” referenced)
• A shorting-adjacent trader example (names distorted by subtitles; Michael Burry mentioned explicitly)

Institutional references

• Goldman Sachs (VIP basket / desk interpretation)