Understand herding and recency bias — Hinglish notes

6.6.9 · Stock-Market › Factor & Behavioral Finance

Overview

Do sabse powerful behavioral biases jo market anomalies drive karte hain aur systematic mispricing create karte hain: herding (crowd ko follow karna) aur recency bias (recent events ko zyada weight dena). Ye biases bubbles, crashes, momentum effects, aur markets ke fundamental values se zyada shoot karne ko explain karte hain.

Core Concepts

The Mechanics: Ye Biases Exist Kyun Karte Hain

Herding Ki Evolutionary Psychology

Herding exist kyun karta hai?

Savanna par, jab sab bhagte hain toh crowd follow karna = survival
Social learning efficient hai: doosron ko copy karna time bachata hai
Safety in numbers individual predation risk kam karta hai
Humans social creatures ke roop mein evolve hue jahan group cohesion > individual optimization

Markets mein ye kaise manifest hota hai?

Information asymmetry: Tum doosron ki trades observe karte ho lekin unki reasoning nahi
Uncertainty: Jab fundamentals unclear hoon, social proof evidence ban jaati hai
Feedback loops: Herding → price moves → herding ko validate karta hai → aur herding

Herd ko kya todta hai?

Fundamentals eventually reassert hoti hain (earnings, cash flows)
Deep pockets wale contrarian traders (value investors)
External shocks jo narrative change kar dete hain

Recency Bias Ka Cognitive Basis

Recency bias exist kyun karta hai?

Memory accessibility: Recent events retrieve karne mein aasaan hote hain (availability heuristic)
Adaptive recency: Changing environments mein, recent data IS zyada relevant hota hai
Emotional vividness: Recent losses/gains abhi bhi sting/thrill karte hain, expectations ko anchor karte hain
Working memory limits: Full historical distribution hold nahi kar sakte, isliye recent sample use karte hain

Ye decisions ko kaise distort karta hai?

True probability distribution: Decades ke data par based
Perceived distribution: Last 6-12 months se dominated
→ Systematic forecast errors
→ Recent performance par over/under reaction

Mathematical Framework

Formula Herding Intensity Measure

Hum herding ko Cross-Sectional Absolute Deviation (CSAD) model use karke quantify kar sakte hain:

$CSAD_{t} = \frac{1}{N} \sum_{i = 1}^{N} ∣ R_{i, t} - R_{m, t} ∣$ Jahan:

$R_{i, t}$ = stock $i$ ka return time $t$ par
$R_{m, t}$ = market return time $t$ par
$N$ = stocks ki number Rational asset pricing ke under: $CSAD_{t} = α + γ_{1} ∣ R_{m, t} ∣ + γ_{2} R_{m, t}^{2} + ε_{t}$ Expected: $γ_{2} = 0$ (linear relationship) Herding ke saath: $γ_{2} < 0$ (extreme moves ke dauran return dispersion non-linearly decrease hoti hai)

YE KYUN KAAM KARTA HAI:

Normal times mein: stocks individual fundamentals ke basis par move karte hain → high dispersion
Herding ke dauran: stocks fundamentals ko ignore karke saath move karte hain → low dispersion
Non-linear $γ_{2} < 0$ extremes par is dispersion ke collapse ko capture karta hai

DERIVATION:

Rational pricing se shuru karo: $R_{i, t} = α_{i} + β_{i} R_{m, t} + ε_{i, t}$

Market se absolute deviations lo: $∣ R_{i, t} - R_{m, t} ∣ = ∣ α_{i} + (β_{i} - 1) R_{m, t} + ε_{i, t} ∣$

Stocks ke across average karo: $CSAD_{t} \approx a + b ∣ R_{m, t} ∣ + noise$

Lekin herding ke saath, jab $∣ R_{m, t} ∣$ bada hota hai, investors apna $β_{i}$ ignore karke sirf market follow karte hain: $β_{i}^{h er d} \to 1 for all i$

Ye dispersion REDUCE karta hai, negative quadratic term create karta hai.

BAYESIAN UPDATING GONE WRONG SE DERIVATION:

Rational Bayesian: $P (θ ∣ D_{1}, D_{2}, .., D_{T}) \propto P (θ) \prod_{t = 1}^{T} P (D_{t} ∣ θ)$ Sabhi data equally treat hota hai.

Recency-biased Bayesian: $P (θ ∣ D_{1}, .., D_{T}) \propto P (θ) \prod_{t = 1}^{T} P (D_{t} ∣ θ)^{λ^{T - t}}$ Recent data ko higher exponent milta hai → higher weight.

Gaussian returns ke liye, ye upar wale recency-weighted sample mean par lead karta hai.

Figure — Understand herding and recency bias

Worked Examples

Worked example Example 1: 2021 Meme Stock Frenzy Mein Herding Detect Karna

Setup: GameStop (GME) January 2021 mein $20 se$ 483 tak rally karta hai. Hum test karna chahte hain ki herding hui thi ya nahi. Data: GME aur 100 similar small-cap stocks ke daily returns.

Step 1: CSAD Calculate Karo

Day 1: GME +135%, Market +2%
Stock A: +8%, Stock B: +3%, Stock C: +12%, ... (100 stocks)
CSAD₁ = (1/100)[|8-2| + |3-2| + |12-2| + ... ] = 8.5%

YE STEP KYUN? CSAD measure karta hai ki individual stocks crowd se kitna deviate karte hain. High CSAD = stocks independently move kar rahi hain.

Step 2: Normal period comparison

Normal day: Market +1%, CSAD = 3.2%
GME frenzy: Market +2%, CSAD = 1.8%

LOWER KYUN? Herding ke dauran, SAARI stocks GME ke saath move karti hain fundamentals ko ignore karke. Individual differences collapse ho jaate hain.

Step 3: Regression test

CSAD_t = α + γ₁|R_m,t| + γ₂R²_m,t
Normal period: γ₂ = 0.02 (slight positive)
Frenzy period: γ₂ = -0.15 (strongly negative)

NEGATIVE KYUN? Negative quadratic term ka matlab hai: "market move jitna bada hoga, stocks utna zyada saath move karengi" = herding signature.

Conclusion: Herding ka strong evidence hai. Retail investors ek group ke roop mein meme stocks mein pile ho gaye, individual fundamentals ko ignore karte hue.

Worked example Example 2: Tech Stock Expectations Mein Recency Bias (2022 Crash)

Setup: 2020-2021 tech boom (+40% annual returns) ke baad, investors extrapolate karte hain. Phir 2022 aata hai -30% crash ke saath. Recency bias ne forecasts ko kaise distort kiya?

Step 1: Recency-weighted expected return calculate karo Historical data:

2022: -30%
2021: +40%
2020: +45%
2019: +30%
2018: -5%

Rational expectation (equal weights):

E[R] = (1/5)[-30 + 40 + 45 + 30 - 5] = 16% average

Recency-biased (λ=0.7) early 2022 mein (crash se pehle):

w₀ = (1-0.7) × 0.7⁰ = 0.30
w₁ = (1-0.7) × 0.7¹ = 0.21
w₂ = (1-0.7) × 0.7² = 0.147
w₃ = (1-0.7) × 0.7³ = 0.103
w₄ = (1-0.7) × 0.7⁴ = 0.072
...
E^biased[R] = 0.30(45%) + 0.21(40%) + 0.147(30%) + 0.103(-5%) + ... 
           ≈ 38%

YE STEP KYUN? Recent boom years ko older data ka 2-3x weight milta hai, expectations ko inflate karta hai.

Step 2: Post-crash recency bias (late 2022) Ab -30% sabse recent hai:

E^biased[R] = 0.30(-30%) + 0.21(40%) + 0.147(45%) + ...
           ≈ 5%

KYUN? Same bias, opposite direction! Recent crash ab dominate karta hai, excessive pessimism create karta hai.

Step 3: Rational se compare karo

Rational: 16% (unchanged, long-term average par based)
Biased pre-crash: 38% → Over-valuation
Biased post-crash: 5% → Under-valuation

Trading implication: Recency bias fundamental value ke around predictable swings create karta hai. Contrarian strategy: recency-driven moves ko fade karo.

Worked example Example 3: Bitcoin 2017 Bubble Mein Herding + Recency Interaction

Setup: Bitcoin $1, 000 (J an 2017) se$ 19,000 (Dec 2017) tak rise karta hai, phir $3,000 (2018) par crash karta hai. Dono biases ek doosre ko amplify karte hain.

The Feedback Loop: Phase 1: Initial rise ( $1, 000 \to$ 5,000)

Kuch early adopters buy karte hain (fundamentals-based)
Price rise karta hai, news mein aata hai
Recency bias kick karta hai: "Is saal 5x upar hai!"
Zyada investors recent gains dekhte hain, future mein extrapolate karte hain

Phase 2: Herding accelerate karta hai ( $5, 000 \to$ 15,000)

Media coverage intensify hoti hai
Herding shuru hoti hai: "Sab Bitcoin ki baat kar rahe hain, mujhe bhi lena chahiye"
Friends, family, coworkers sab buy kar rahe hain
FOMO dominate karta hai: Fear of missing out > fundamental analysis
Har buyer agla validate karta hai: "Dekho, abhi bhi upar ja raha hai!"

Phase 3: Mania ( $15, 000 \to$ 19,000)

Recency bias extreme: 6-month return = +400%
λ=0.6 wale investors: E[R_annual] ≈ 300% (absurd)
Herding complete: Google Trends for "Bitcoin" all-time high par
Grandmothers Bitcoin kaise khareedein pooch rahi hain = classic top signal

Phase 4: Crash ( $19, 000 \to$ 3,000)

Kuch whales sell karte hain, price 20% drop karta hai
Recency flip negative hota hai: Recent return = -20%
Herding reverse hoti hai: "Sab sell kar rahe hain, mujhe bhi karna chahiye"
Margin calls aur selling force karte hain
Naya recency-biased expectation: "Ye zero jaayega"
Herding downside accelerate karta hai: -84% total

Interaction quantify karna:

Herding intensity (CSAD γ₂): 
 Normal: -0.05
 Bubble phase: -0.35 (7x stronger)
Recency weight (λ):
 Normal investor: 0.85
 During bubble: 0.55 (extreme recency)
Combined effect: 7× 1.5 ≈ 10x amplification

YE KYUN MATTER KARTA HAI: Aise hi saare bubbles kaam karte hain. Recency bias over-extrapolation cause karta hai. Herding initial move ko amplify karta hai. Feedback loop tab tak continue karta hai jab tak fundamentals reassert na ho jaayein ya greater fool khatam na ho jaaye.

Common Mistakes

Common mistake Mistake 1: "Herding Hamesha Irrational Hai"

Galat intuition: "Crowd follow karna dumb hai. Smart investors independently sochte hain."

Sahi kyun lagta hai: Hum Buffett jaison contrarians celebrate karte hain. Media ko "Main sahi tha jab baaki sab galat the" wali stories pasand hain.

Sach: Herding kaafi situations mein RATIONAL ho sakti hai:

Asymmetric info ke saath information cascades: Agar doosron ke paas private information hai, unhe mimic karna Bayesian-optimal hai
Career risk: Fund managers akele galti karne se bachne ke liye herd karte hain ("IBM kharidne par koi fire nahi hota")
Coordination games: Jab value doosron ke actions par depend kare (jaise bank runs), herding Nash equilibrium hai

Sahi tarike se kaise sochein:

Rational informational herding aur emotional contagion mein distinguish karo
Poochho: "Kya mere paas woh information hai jo doosron ke paas nahi?" Agar nahi, toh kuch herding justified hai
PROBLEM fundamentals ko ignore karne wali herding hai, har herding nahi

Example: Agar 5 expert investors jinhein tum respect karte ho sab ek stock buy karte hain, toh unhe follow karna (apne analysis ke bina) necessarily irrational nahi hai—unhe tumse zyada pata ho sakta hai.

Common mistake Mistake 2: "Recency Bias Ka Matlab Hai Recent Data Useless Hai"

Galat intuition: "Kyunki recency bias exist karta hai, mujhe recent events ignore karke sirf long-term averages dekhne chahiye."

Sahi kyun lagta hai: Hume bola jaata hai long-term sochao, recent performance chase mat karo. "This time is NOT different."

Sach: Recent data zyada informative HO SAKTA HAI jab duniya structurally change ho gayi ho:

Regime shifts: Pre-2008 vs. post-2008 monetary policy
Technology disruption: 2010 vs. 2020 mein Netflix
Company lifecycle: Startup vs. mature company

Bias recent data ko OVER-weight karna hai, na ki recent data ignore karna chahiye.

The fix:

Poochho: "Kya underlying process change hua hai?"
Agar haan: Recent data ko zyada weight do (lekin phir bhi 100% nahi)
Agar nahi: Recency bias avoid karne ke liye longer history use karo
Structural breaks ke liye statistical tests use karo

Example: Tesla ki 2018-2020 growth rate 2012-2014 se zyada relevant IS hai kyunki production scale hui aur battery costs girein. Lekin 2020 growth ko forever extrapolate karna recency bias hoga.

Common mistake Mistake 3: "Main Opposite Karke In Biases Ko Exploit Kar Sakta Hoon"

Galat intuition: "Sab tech stocks mein herd kar rahe hain, toh main unhe sab short karunga. Contrarian = profit."

Sahi kyun lagta hai: Behavioral finance sikhata hai biases mispricings create karte hain. Biases exploit karo → alpha!

Sach: Biases correct hone se pehle persist aur worsen ho sakti hain:

Limits to arbitrage: Bubbles short karna tumhe bankrupt kar sakta hai sahi hone se pehle ("Markets irrational rahein zyada der tak jitna tum solvent reh sako" - Keynes)
Bias timing: Herding KHATAM kab hogi? Koi precisely nahi jaanta
Tum galat bhi ho sakte ho: Shayad "herding" actually informed buying hai

The fix:

Herd ke DAURAN herding se mat lado—ye reverse mein falling knife pakadna hai
Early reversal signs ka wait karo (momentum breakdown, volume exhaustion)
Early hone ke liye position size karo (contrarian par all-in mat jao)
Defined risk ke liye options use karo
Fundamental anchors ke saath combine karo: sirf tab herding fade karo jab price >> value ho

Example: Tesla ko $200 (2020) p a r h er d in g k i w aj ah ses h or t k a r na t u mh e ba r ba d k a r d e t ak y u nk i y e$ 1,200 (split-adjusted) tak gayi. Better: momentum break + valuation catalyst ka wait karo.

Active Recall Practice

Recall Feynman Technique: Ek 12-Saal-Ke-Bache Ko Explain Karo

Herding: Socho tum school mein ho aur suddenly sab playground ki taraf bhagte hain. Tum nahi jaante kyun, lekin tum bhi bhagte ho kyunki "kuch zaroor ho raha hai!" That's herding. Stock market mein, jab bahut log ek stock buy karte hain, doosre price upar jaate dekh sochte hain "mujhe bhi kharidna chahiye!" Lekin woh isliye nahi kharid rahe kyunki company achi hai—woh isliye kharid rahe hain kyunki baaki sab kar rahe hain. Eventually, jab logon ko pata chalta hai koi real reason nahi tha buy karne ka, sab ek saath sell karne ki koshish karte hain, aur price crash kar jaati hai.

Recency Bias: Socho tum ek coin 100 baar flip karte ho. 70 baar heads aur 30 baar tails aata hai. Lekin last 10 flips mein, 9 baar heads tha. Ab tumhe next flip par bet lagana hai. Tumhara brain chillaata hai "HEADS!" kyunki tumne abhi wahi dekha. Lekin coin abhi bhi 50/50 hai! Recency bias ka matlab hai tumhara brain JO ABHI HUA usse zyada dhyan deta hai aur badi picture bhool jaata hai. Stocks mein, agar kisi company ki price last 3 months se upar gayi, log sochte hain ye hamesha upar jaayegi, bhool jaate hain ki 10 saalon mein ye upar bhi jaati hai AUR neeche bhi.

Connections

Momentum and reversal effects - Herding momentum create karta hai; recency bias explain karta hai kyun ye overshoot karta hai
Market bubles and crashes - Bubble formation aur collapse ke liye dono biases zaroori hain
Behavioral portfolio theory - Explain karta hai kyun investors diversify nahi karte (popular stocks mein herd karte hain)
Limits to arbitrage - Kyun smart money hamesha herding-driven mispricings correct nahi kar sakti
Information cascades - Rational herding ki theoretical foundation
Prospect theory - Recency bias loss aversion amplify karta hai (recent losses zyada painful)
Overconfidence bias - Recency ke saath interact karta hai: recent success → overconfidence → herding
Social proof and FOMO - Herding ke underlying psychological mechanisms
Contrarian investing strategies - Practical application: extreme herding + recency fade karo
Market efficiency debate - Ye biases strong-form efficiency ke against key evidence hain

#flashcards/stock-market

What is herding behavior in financial markets? :: Financial markets mein investors ki tendency jo ek bade group ke actions ko mimic karte hain, isliye buy karte hain kyunki doosre buy kar rahe hain ya isliye sell karte hain kyunki doosre sell kar rahe hain, fundamental analysis se independent hokar. Informational cascades aur emotional contagion create karta hai.

What is recency bias?

Recent events ko zyada weight dene aur long-term base rates ko kam weight dene ka cognitive error jab expectations form karte hain. Recent information zyada vivid hoti hai aur yaad karna aasaan hota hai, jo probability estimates ko distort karti hai.

How do we measure herding intensity quantitatively?

Cross-Sectional Absolute Deviation (CSAD) use karke. CSAD_t = α + γ₁|R_m,t| + γ₂R²_m,t regress karo. Agar γ₂ < 0 (negative quadratic term) hai, toh herding indicate hoti hai—extreme market moves ke dauran return dispersion non-linearly decrease hoti hai.

What does γ₂ < 0 in CSAD regression tell us?

Herding present hai. Negative γ₂ ka matlab hai ki bade market moves ke dauran, individual stock returns diverge karne ki bajaye converge hote hain (low dispersion), jo indicate karta hai ki investors individual fundamentals ki jagah crowd follow kar rahe hain.

How do we model recency-biased expectations mathematically?

E_t^biased[R_{t+1}] = Σ w_s R_{t-s} jahan w_s = (1-λ)λ^s. Recency bias ke saath, λ < 1 (typically 0.6-0.8), recent observations ko door wale observations se exponentially higher weight milta hai.

What does λ = 0.7 mean in recency-weighted expectations?

Recent data ko exponential decay ke saath higher weight milta hai. Most recent period ko 30% weight milta hai, ek period pehle ko 21%, do period pehle ko 14.7%, etc. Last year teen saal pehle se 2-3x zyada matter karta hai.

Why is herding not always irrational?

(1) Information cascades: doosron ke paas private information ho sakti hai, isliye unhe mimic karna Bayesian-optimal hai. (2) Career risk: fund managers akele galtiyan karne se bachne ke liye herd karte hain. (3) Coordination games: jab value doosron ke actions par depend kare, herding Nash equilibrium hai.

What's the key interaction between herding and recency bias in bubbles?

Positive feedback loop: (1) Recent price rise → recency bias expectations inflate karta hai. (2) High expectations → zyada log aate hain toh herding. (3) Herding → higher prices. (4) Loop repeat hota hai jab tak fundamentals reassert na ho jaayein. Bubbles mein 10x amplification create karta hai.

When is recent data actually MORE informative than long-term averages?

Structural regime shifts ke dauran: monetary policy changes, technology disruption, company lifecycle transitions. Bias recent data ko OVER-weight karna hai, na ki recent data irrelevant hai. Structural breaks ke liye tests use karo.

Why is "fade the herd" not always profitable?

(1) Limits to arbitrage: bubbles short karna correction se pehle bankrupt kar sakta hai. (2) Bias timing: herding expected se zyada der persist kar sakti hai. (3) Tum galat bhi ho sakte ho: apparent herding actually informed buying ho sakti hai. Markets tumhare solvent rehne se zyada der irrational reh sakta hai.

What are the three types of herding?

(1) Informational cascades: "Unhe kuch pata hai jo mujhe nahi." (2) Reputational herding: "Mujhe fire nahi karenge agar main wohi karun jo sab karte hain." (3) Emotional contagion: Fear aur greed socially networks mein spread hote hain.

How did herding manifest in the 2021 meme stock frenzy?

CSAD extreme moves ke dauran collapse ho gayi—γ₂ strongly negative ho gayi (-0.15 vs normal +0.02). Saari stocks individual fundamentals ko ignore karke GME ke saath move karti rahein, kyunki retail investors social media se driven hokar ek group ke roop mein pile ho gaye.

What was the recency bias effect on tech expectations 2021→2022?

Pre-crash: λ=0.7 weighting ne expected returns 38% tak inflate kar diye (rational 16% ke vs), overvaluation cause karte hue. Post-crash: same bias flip ho gayi negative, expectations 5% tak deflate ho gayi, undervaluation cause karte hue. Same bias, opposite extremes.