Week 8: Bubbles and Seasonality

1. Introduction: two kinds of patterns in prices

The lecture examines how mass psychology and patterns that keep recurring produce price anomalies, in two parts. Part one covers bubbles: their definition, how they arise, well-known cases from 1637 to 2021 (and the later AI boom), the psychology driving them (narratives, herding, FOMO) and the methods for spotting them. Part two covers seasonality, or calendar anomalies: the January effect, the turn-of-the-month effect, the day-of-the-week effect, the Halloween effect and the holiday effect, together with behavioral and microstructure explanations.

The opening slide gives the shared message: calendar anomalies and bubbles both show that markets fall short of full efficiency. The Efficient Market Hypothesis (EMH) says prices contain all available information, so they ought to stay near fundamental value and returns ought not to be forecastable from something as trivial as the calendar. Each phenomenon covered here violates one of these two claims.

2. What is a bubble?

The slides list five main characteristics:

• Rapid price increases without corresponding fundamental support. Prices climb much faster than earnings, dividends or rents.
• Stories or narratives that justify a "new paradigm". People say "this time is different": the old valuation rules supposedly no longer apply.
• Excessive participation of non-professionals. Taxi drivers, students and relatives start talking about the asset and buying it.
• High leverage and margin trading. Investors borrow to buy more, which pushes prices up further but makes the system fragile.
• A sharp collapse (crash) when sentiment turns. The fall is usually much faster than the rise.

Robert Shiller (2000), in Irrational Exuberance, defines bubbles as "episodes of social epidemics of optimism". The word "epidemic" is chosen on purpose: optimism spreads from person to person like a virus, and the price rise itself is the evidence that convinces the next group of people.

The psychological cycle of a bubble

The slides show the well-known "Market Cycle Investing Guide", drawn on top of the NASDAQ price path from 1998 to 2004 (rising from roughly 1,000 to above 4,500 in 2000 and then falling back). It labels thirteen emotional stages:

• Rising phase: 1 Hope, 2 Optimism, 3 Belief, 4 Thrill, 5 Euphoria (the peak).
• Falling phase: 6 Complacency, 7 Anxiety, 8 Denial, 9 Panic, 10 Capitulation, 11 Anger, 12 Depression (the bottom).
• New cycle: 13 Disbelief (the recovery starts but nobody trusts it).

The slide summarises this as two emotional phases: excitement, greed and euphoria on the way up, and anxiety, denial and panic on the way down. The point for investors is that emotions run opposite to the right decision: people feel most confident exactly at the top (euphoria), when expected returns are lowest, and feel worst at the bottom (depression, disbelief), when expected returns are highest. "Denial" links to earlier weeks: after the first drop, investors hold on because selling would mean realising a loss (the disposition effect and loss aversion from prospect theory).

3. Historical bubbles

Tulip Mania (1637)

Tulip Mania in the Netherlands is described as the first recorded financial bubble:

• There was a mania for rare tulip varieties.
• Prices rose by 20x to 50x.
• Many ordinary individuals took part (merchants, craftsmen), not only professional traders.
• The collapse happened within a few weeks.
• At the peak, a single bulb cost about 10 times the annual salary of a craftsman.

The slide presents this as a case of "narrative contagion": the story mattered more than the value. A tulip bulb yields flowers, not cash flows, so nothing rational could justify paying ten years of wages for one. What held the price up was only the expectation that a later buyer would pay more still.

The Dot-Com bubble (1999 to 2000)

• Massive optimism around the Internet.
• IPOs (initial public offerings) of firms without profits, but with a "story".
• The NASDAQ rose about +400% within 4 years.
• It then collapsed by about 78% in 2000 to 2002.

The second slide shows the NASDAQ-100 index from 1995 to early 2000: a gradual rise until 1998, then an almost vertical climb to about 4,400 to 4,500 in March 2000, far above its 200-day moving average. A price that moves far above its own trend line is a typical visual sign of a bubble.

Crypto bubbles: 2017 and 2021

For Bitcoin and altcoins the slides list four features:

• Massive retail inflow, driven by social media.
• Excessive leverage on trading platforms (Binance, FTX).
• A "get rich quick" narrative.
• A rapid decline when sentiment turned.

Two tables (from "Bitcoinmania: A Ticking Time Bomb Waiting to Explode", 2021, based on CoinMarketCap data) illustrate how large the 2021 boom was and how quickly it collapsed.

Notes on the tables: the $2.02 trillion of the top 10 coins was 80.45% of the combined market cap of all 9,651 coins ($2.51 trillion). By 23 May 2021 the total crypto market cap was $1.24 trillion, of which $590.2 billion was Bitcoin. In about fourteen months the top 10 coins gained more than 14 times their value, and then lost half of it in less than two weeks.

Dogecoin is the clearest narrative asset in the table: a coin started as a joke rose more than 300 times (30,578.5%), largely because of social media attention. Tether is the control case: it is a stablecoin designed to stay at $1, so its price did not move at all.

4. Why are bubbles created?

The slides give six behavioral reasons:

• Narrative contagion (Shiller): a success story spreads from person to person.
• Herding: investors imitate one another. It feels safer to be wrong together than to be wrong alone, and others' buying is taken as information.
• FOMO (fear of missing out): the fear of "being left out" while friends get rich. This is regret aversion: the pain of missing a gain that others received.
• Overconfidence: investors overestimate their abilities and believe they will get out before the crash.
• Attention-driven buying: media coverage draws attention, and attention brings inflows. Retail investors mostly buy what they notice.
• Low interest rates: cheap money encourages risk-taking and leverage. When safe assets pay almost nothing, investors "search for yield" in risky assets.

Note that the last reason is not purely psychological: it is a monetary condition. Bubbles usually combine a real economic trigger with a psychological amplifier, a theme that returns in the housing bubble below.

5. How do we detect a bubble?

Main indicators listed on the slides:

• Extreme P/E and P/S ratios (price to earnings, price to sales), a disconnect from fundamentals.
• Very high trading volume.
• Excessively high IPO valuations.
• High leverage (for example record margin debt).
• Google searches for "how to buy...": a sign that new, inexperienced buyers are entering.
• Social media clustering (Twitter and Reddit hype).

The slide concludes: a bubble is more of a social than a financial phenomenon. Half of the indicators are not about prices at all but about who is buying and how much they talk about it.

6. Valuation tools: CAPE, P/E, P/B and Tobin's Q

The Shiller CAPE

• Aim: to spot periods of overvaluation or undervaluation.
• Developed by Campbell & Shiller (1988, 1998).
• Cyclical volatility is filtered out by smoothing earnings.

Why 10 years? Earnings move strongly with the cycle: they slump in recessions and balloon in booms. Averaging over 10 years "wipes out" the business cycle, so the denominator captures normal earning power. "Real" means the past earnings are corrected for inflation, putting earnings from a decade ago in today's money. The slide presents CAPE as the best-known tool for spotting bubbles in the stock market.

What does CAPE mean for future returns?

Campbell & Shiller (1998) found that CAPE has a strong negative relationship with the following 10-year returns of the S&P 500.

• High CAPE: investors have overestimated earnings, so future returns are low.
• Low CAPE: undervaluation, so future returns are higher.
• CAPE works as a "market thermometer".

Put briefly: a bigger bubble today means weaker returns over the following ten years. The logic is basic arithmetic. Paying 40 times normal earnings gives an earnings yield (its inverse, 1/40) of just 2.5% a year. Paying 10 times gives an earnings yield of 10%. Across ten years, this gap in starting yield is what mostly drives returns.

The chart "CAPE ratio over time" (1880 to today) shows the long-run average around 15 to 17, with peaks at about 32 just before the 1929 crash, an all-time high of about 44 at the dot-com peak in 1999 to 2000, lows near 5 around 1920 and about 7 in the early 1980s, and a current value of about 40 (40.11 on the slide), close to the dot-com level.

CAPE versus the classic P/E

• P/E is based on current (or forecast) earnings. CAPE is based on smoothed real earnings over 10 years.

Advantages of CAPE over P/E:

• Not affected by cyclical earnings fluctuations (recession or expansion).
• Does not depend on aggressive accounting practices by firms in a single year.
• Gives a more stable estimate of "long-run valuation".

Disadvantages of P/E:

• It may look "cheap" after a price drop, or misleading because E itself collapses (see the recession example above).
• Forward P/E (price divided by next year's expected earnings) depends on analyst forecasts, which are biased and usually too optimistic.

Conclusion on the slide: CAPE is more appropriate for long-term market valuation.

Price-to-Book and Tobin's Q

Price-to-Book (P/B) compares market capitalization with the book value of equity (assets minus liabilities as recorded in the balance sheet). It is most suitable for banks and industrial firms, whose assets are mostly tangible or financial and recorded close to their value.

If Q is above 1, the market values the firm at more than it would cost to rebuild it from scratch: a sign of overvaluation (or of valuable intangibles). If Q is below 1, the firm is worth less than its assets would cost to replace: undervaluation. Q is theoretically a fundamental indicator, but difficult to compute, because nobody knows the replacement cost of brands, software or patents.

CAPE versus the others, and which indicator is reliable

• CAPE is a "cleaner" valuation measure because it is based on earnings, not on book values or replacement cost.
• P/B and Q help, but they are less stable and, by construction, exposed to shifts in accounting rules.

(The slide also contains a "sensitivity" column that is partly cut off; what can be seen indicates that forward P/E reacts strongly to analyst bias.) Conclusion: CAPE predicts returns over several years (5 to 10+ years) best, whereas P/E and P/B work better for stock selection (picking which stock) than for market timing (choosing when to be invested).

7. Theories of bubble formation

Bubbles as socio-psychological phenomena (Shiller 2000, 2019)

In Irrational Exuberance (2000) and Narrative Economics (2019), Shiller argues that bubbles are not just overvaluations but social narratives that spread:

• What persuades investors is not data but stories.
• A bubble is an "epidemic" of a viral narrative.
• The media amplify and recycle the stories, which speeds up their spread.
• The stories create expectations, expectations lead to real investments, and investments push prices up.

Central sentence: first comes the story, and only then the price. That turns the efficient-markets logic upside down, in which fundamental news drives prices. In this case a story drives prices, and the price rise in turn makes the story more credible, so a feedback loop forms.

The Minsky mechanism: from confidence to crisis

Hyman Minsky's Financial Instability Hypothesis describes three phases of leverage:

1. Hedge finance: prudent financing, low leverage. Income covers both interest and repayment of principal.
2. Speculative finance: investors can pay the interest but do not reduce the principal; they need to roll over (refinance) their debt.
3. Ponzi finance: investors rely on rising prices to service their debt. Their income does not even cover interest.

What sets off the crash? A modest turn in the trend produces margin calls (lenders ask for more collateral), these produce forced sales, and the sales drive prices lower still. The herd changes course and panic takes hold of the market.

Minsky's famous phrase is "stability is destabilizing": when everything goes well for a long time, investors and lenders become comfortable and increase risk too much, moving from hedge to Ponzi finance without noticing.

How a bubble forms: a 6-stage mechanism

1. Displacement: a new technology (Internet, crypto, AI) or a political change creates new profit opportunities.
2. Credit expansion: cheap borrowing (low rates) and high liquidity.
3. Euphoria: the media create wealth stories; the number of investors rises steeply.
4. Mania: everyone buys; excessive leverage.
5. Distress: a small correction creates concern.
6. Revulsion: panic, collapse, "mass exit".

Main point: when all around them are getting rich, investors stop noticing risk. The 6-stage model (usually associated with Minsky and Kindleberger) works well as a checklist: for each episode, ask what the displacement was, what source of credit fed it, and what set off the distress.

Why bubbles repeat themselves

Despite 400 years of history, bubbles keep coming back. Reasons:

• Learning myopia: the previous crisis is soon forgotten.
• Human behavior: human psychology stays the same, and investors overreact to stories.
• Weak memory and recency bias: investors concentrate on recent returns and project them forward.
• Technological change differs: every generation gets its own "new technology", so the story always seems new.

Conclusion: bubbles are a product of human psychology, not of economic models.

8. Modern examples: Bitcoin, AI and housing

Bitcoin as a narrative bubble

Following Shiller (2019), Bitcoin is a classic narrative-driven bubble because:

• With no classical fundamentals such as cash flows or earnings, only stories can move its price; Bitcoin has no CAPE or P/E.
• The story of a "decentralized future, freedom from banks" went viral.
• Social media (Twitter, Reddit, YouTube) magnified both the narrative and FOMO.
• A rally led by retail investors switched quickly to panic once sentiment turned.

Slide message: narratives are stronger than valuation.

The AI bubble (2023 to 2024): the latest major narrative

• Explosive publicity around ChatGPT (late 2022 to early 2023).
• Stocks such as NVIDIA, AMD, SuperMicro and Palantir soared.
• FOMO: massive retail and institutional flows into AI-related stocks.
• The media reinforced the narrative "AI changes everything".

The slide frames AI as a Minsky-type "displacement event": stage 1 of the 6-stage mechanism. The NVIDIA chart shows its market capitalization rising from a low of about $0.35 trillion at the end of 2022 to a high of about $5 trillion in late 2025, around $4.3 trillion at the time of the slide: more than a twelvefold rise in under three years. The slide labels this a bubble. A fair academic view is that AI is a case where real fundamentals (NVIDIA's earnings did grow strongly) and narrative amplification are mixed, which makes it hard to say how much of the price is bubble. That mix is exactly the next example.

The housing bubble (2002 to 2007): behavior amplified the fundamentals

What caused it? A mixture of fundamentals and a behavioral collective illusion.

Fundamental drivers:

• Very low interest rates (Federal Reserve policy).
• Loose lending standards (subprime mortgages to borrowers with weak credit).
• Securitization: mortgages were bundled into MBS (mortgage-backed securities) and CDOs (collateralized debt obligations) and sold on, so the original lender no longer carried the risk.

Behavioral amplification:

• Representativeness: "house prices always rise", a belief built on a short recent history that seemed representative of the future.
• Moral hazard: brokers were paid per loan and passed the risk on, so when issuing loans they had no incentive to check creditworthiness.
• Overconfidence: buyers and investors taking on several loans without any savings.
• Herding: mass buying of property and "flipping" (buying in order to resell quickly).

The slide concludes: the 2008 crisis had psychological roots as well as economic ones.

9. Behavioral trading inside bubbles

Types of investors in a bubble:

• Noise traders: trade on hype and emotion instead of information.
• Smart money: aware it is a bubble, they try to get out before the rest. For a while their buying can even push the bubble higher (they ride it instead of fighting it).
• True believers: they truly believe the story.
• Momentum traders: they follow the trend and buy because prices have risen.

Behavioral strategies:

• Ride-the-bubble strategy: get in early and get out once volatility starts.
• Contrarian at the peak: short-sell when sentiment is extreme.
• Narrative monitoring: use NLP (natural language processing, text analysis of news and social media) to detect the "narrative peak".

The slide's rule: a bubble survives longer than you think it will and bursts more quickly than you can picture. This ties in with the limits of arbitrage: a rational short-seller may judge the value correctly and still go broke, because the bubble can keep inflating for months or even years before it bursts.

10. Seasonality: calendar anomalies

Key anomalies: the January effect, the turn-of-the-month effect, the day-of-the-week effect, the Halloween effect ("Sell in May and go away") and the holiday effect. Because they are predictable and persistent, they contradict the EMH. If it were common knowledge that some dates bring higher returns, rational traders would buy shortly before them, pulling the gains forward until the pattern vanished.

A look at the data: "September is the worst month for the S&P 500"

A heat map on the slides shows monthly S&P 500 returns from 2008 to 2025. The 17-year averages per month are:

September is the only month with a clearly negative average (-1.16%), with large losses in 2008 (-9.08%), 2011 (-7.18%), 2022 (-9.34%) and 2023 (-4.87%). Two lessons follow. First, monthly patterns exist even in the most liquid index in the world. Second, notice that January is slightly negative (-0.11%) for the S&P 500 in this recent period: the January effect is mainly a small-cap effect (see Keim below), and it is weaker in large stocks and in recent decades. Also be careful: 17 observations per month is a small sample, and one extreme year (like 2008) can drive an average.

11. The January effect

What was observed: returns are higher in January, above all for small caps, frequently by 1 to 3% in that month alone. Possible explanations:

• Tax-loss selling (December, January): investors sell losing stocks before year-end to realise a capital loss that lowers their tax bill, then buy back in January.
• Window dressing by fund managers: before the year-end report, managers sell embarrassing losers and buy recent winners so the portfolio looks good, then return to riskier small stocks in January.
• Behavioral: optimism is renewed as the new year begins.

Rozeff & Kinney (1976): the discovery

Published in the Journal of Financial Economics under the title "Capital Market Seasonality: The Case of Stock Returns".

• The first systematic evidence that returns in January are statistically significantly higher.
• Data spanning many decades: 1904 to 1974.
• In the US, January returns average 3.5 to 5%, against 0.7% in the other months.
• The anomaly cannot be explained by risk, beta or market structure.

This paper established the idea that markets have calendar anomalies. Their sample was an equal-weighted index, which gives small stocks a large weight; that detail is the bridge to Keim.

Keim (1983): a small-cap phenomenon

Published in the Journal of Financial Economics under the title "Size-Related Anomalies and Stock Return Seasonality".

• Small-cap firms show a far stronger anomaly.
• In January, small caps earn returns 6 to 10 times higher than in other months.
• Most of the annual excess return on small caps comes in January alone.
• It is tied to December tax-loss selling, liquidity constraints and behavioral biases (optimism, the year-end reset).

Keim established the January effect as a size anomaly. This matters for asset pricing: the "small-firm premium" of Banz and of Fama-French's SMB factor is concentrated in a few weeks of the year, which is hard to reconcile with a risk premium that should be earned evenly over time.

Chen & Singal (2004): the modern interpretation

Published in the Journal of Financial Markets as "All About the January Effect". Two forces combine to produce the January effect:

1. Tax-loss selling (December): as the year closes, investors dump their "losers", driving those prices lower.
2. Disposition effect (behavioral): during the year, investors avoid selling stocks with losses (loss aversion, regret). So the selling of losers is postponed and clusters in December, followed by a rebound in January.

Empirical findings:

• The biggest Q4 losers post the largest January gains.
• Even with reduced tax incentives the effect remains.
• Risk and market microstructure do not explain it.

According to the slide, this paper settles the behavioral account of the January effect.

12. The Halloween effect: "Sell in May and go away"

Observation: the November to April period earns far more than the May to October period. The slide's bar chart (S&P 500, Bloomberg data):

So on average, almost all of the yearly equity return arrives in the winter half-year. The gap is a bit smaller in the more recent sample, but it is still large. Explanations:

• Summer holidays and thin liquidity.
• Lower institutional participation because of seasonality (fewer traders at their desks during summer).
• In autumn, more volatility and political uncertainty.
• Early in the year, investors underreact behaviorally to corporate news.

13. The turn-of-the-month (TOM) effect

• A stable finding worldwide.
• Linked to capital flows into funds, pensions and payroll cycles.
• Behavioral: anchoring and adjustment, and institutional herding.

Key papers

Explanation: institutional flows. Pensions, salaries and payroll create liquidity around the turn of the month, and that money is invested in the market in the first days of the new month.

Evidence from an emerging market: BIST100

The slide shows mean daily returns of the Turkish BIST100 index for each day from -10 to +10 around the turn of the month (Kayacetin & Lekpek, 2016, Finance Research Letters). The pattern is clear: on most ordinary days the mean daily return is between about -0.1% and +0.2%, but on day -1 it is about 0.5%, on days +1 and +2 about 0.4 to 0.45%, and on day -2 and day +4 about 0.25 to 0.35%. The return is concentrated in a short window around the month change, also in an emerging market.

The TOM strategy

Strategy: hold the S&P 500 from day -3 before the end of the month through day +3; stay out of the market, or sit in a risk-free asset (T-bills), for the rest of the month. Results (Hensel & Ziemba 1996, Kunkel et al. 2003): TOM days deliver 70 to 90% of the S&P 500's long-run gains; the TOM strategy beats buy-and-hold on Sharpe ratio, even net of trading costs; and the effect is found in the US, Europe, Asia and emerging markets.

Behavioral explanation: institutions herding as the month starts; a short-lived boost in liquidity together with a "relief" mood; and arbitrageurs paying little attention on the days in between. The slide points out that TOM is one of only a handful of seasonal trades showing long-term persistence.

14. Why do calendar anomalies exist?

The paradox: seasonality anomalies are predictable, so they violate the EMH. Behavioral explanations:

• Prospect theory: sentiment differs at the beginning and the end of periods, because gains and losses are judged relative to a reference point (for example the price at the start of the year).
• Mental accounting: investors "close their books" at calendar points (year-end, quarter-end) and open a new account afterwards.
• Overreaction and underreaction to earnings and announcements, which arrive in seasonal waves.
• Herding: institutional reshuffling of portfolios in specific periods.
• Attention cycles: interest in markets varies by month.

Slide conclusion: seasonality is psychological, not fundamental. Firms do not earn more cash in January or on the last day of the month, so the patterns must come from how and when investors trade.

15. The day-of-the-week (weekend) effect

Observation: Monday returns are negative on average, Friday returns are more positive, and this shows up across many countries and markets. Behavioral accounts: a gloomy Monday mood (the mood hypothesis) and expectations being revised after the weekend.

The chart "Returns by day of the week, 1927 to 2001" (from Aswath Damodaran) shows that Monday is the only day with a negative average daily return (about -0.12%), while Tuesday to Friday are positive (Wednesday and Friday highest, around 0.09 to 0.11%). The line for the share of days with a gain is lowest on Monday (about 46%) and about 51 to 54% on the other days.

Theory and history: key papers

• Cross (1973): the first to notice it: Mondays negative, Fridays positive.
• French (1980, JFE): established the weekend effect. If returns accrued in calendar time, Monday (spanning three days, from Friday's close to Monday's close) ought to show the highest return; in fact it shows the lowest.
• Gibbons & Hess (1981): statistical strength for the US over 1962 to 1978.
• Jaffe & Westerfield (1985): confirmed the effect in international markets.
• Lakonishok & Maberly (1990): a behavioral explanation based on how retail investors trade.

Summary sentence: Monday "pays" for the weekend, Friday reflects a better mood.

Behavioral and microstructure explanations

Behavioral mechanisms:

• Mood hypothesis (Kamstra, Kramer & Levi, 2003): people feel worse on Mondays, which lowers their appetite for risk.
• Investor inattention (DellaVigna & Pollet, 2009): investors pay less attention on Fridays, so news released on Friday gets an underreaction; the slide links this to positive returns on Monday as prices catch up.
• Disposition and weekend information: investors process news during the weekend, and adjust their positions on Monday.
• Retail trading patterns (Lakonishok & Maberly, 1990): since retail investors make up their minds over the weekend, they sell more on Mondays, which pushes prices down.

Microstructure mechanisms (market microstructure means the mechanics of trading: spreads, liquidity, order flow):

• Information builds up while markets are closed and enters prices all at once.
• Liquidity is lower.
• Bid-ask spreads are wider on Mondays.

Weekend effect and the mood hypothesis

Why does Monday have lower returns? According to the mood hypothesis, psychology experiments show lower mood on Mondays; lower mood means pessimism and lower demand for risk, which is reflected in stock prices. Empirically, Monday has the highest probability of a negative return and Friday the highest probability of a positive one. Mood is reflected in volatility and returns.

16. The holiday effect

Observation: returns are unusually high on the trading days just before holidays, an effect found in over 20 countries. Behavioral accounts: investors are in a better mood (a positive mood effect), arbitrageurs pay less attention, and there is a habit of "window dressing" ahead of month-end or quarter-end. Message: investor mood affects prices.

17. Earnings announcements and post-earnings announcement drift (PEAD)

Stocks show a systematic underreaction to earnings. After a positive earnings surprise, the price jumps on the announcement day but then keeps drifting up for weeks; after a negative surprise, it keeps drifting down.

Mechanism: information is absorbed with a delay; attention is low (inattention) in "noisy" periods; and the drift is stronger during periods of mass announcements (Q1, Q4), when lots of companies report together and investors cannot digest everything. That explains its place in the seasonality lecture: earnings seasons are fixed on the calendar. The slide describes PEAD as one of the most persistent behavioral anomalies.

18. The tax-loss selling effect

As the year ends, investors sell their "losers" for tax reasons; early in January those stocks bounce back. Once more the behavioral story is the disposition effect: during the year people hold on to losers, and since the tax deadline makes them sell in December, the selling comes all at once and is followed by a rebound. This feeds the January effect, above all in small caps.

19. Seasonality in 30+ international markets

• These effects are global (US, EU, Asia, emerging markets).
• In small markets dominated by value stocks, the January effect is stronger.
• Where institutional trading is low (more retail, more driven by mood), the holiday effect is stronger.
• Since 2000 the day-of-the-week effect has become weaker, though it is still there.

The weakening after publication and after cheaper trading is what the EMH would predict: once an anomaly is known and easy to trade, arbitrage reduces it. The fact that most of these effects did not fully disappear shows that limits to arbitrage and persistent behavior keep them alive.

20. Seasonal strategies

A seasonal strategy exploits predictable calendar-based patterns (January, turn-of-the-month, holiday). They work because of behavioral biases (attention, mood, herding), institutional flows (pension and mutual funds) and market microstructure (liquidity cycles). The question is whether they truly deliver "alpha": a return above what is justified by their risk, after costs.

Although labelled anomalous, the January trade ranks among the most familiar and most stable seasonal trades.

21. Mood cycles and institutional inflow cycles

Mood cycles (Kamstra, Kramer & Levi, 2003): months with less daylight lead to seasonal depression (seasonal affective disorder) and lower risk appetite; months with more daylight lead to better mood and higher returns. Empirically, there is a positive relation between daylight hours and returns, and the effect is strongest in Scandinavian markets, where the difference in daylight between seasons is largest. This is a nice test design: the "treatment" (daylight) is completely unrelated to firm fundamentals, so a link with returns points to psychology.

Cycles of institutional inflows: pension contributions and payroll bring inflows early in the month, and mutual funds rebalance their portfolios at calendar points (month-end, quarter-end, year-end). These flows are not psychological but mechanical, and they create predictable buying pressure.

22. Comparison of seasonality effects

23. How bubbles and seasonality interact

Bubbles often show seasonal escalation. Examples on the slide: in the dot-com bubble, inflows peaked in Q1; in crypto, prices rose strongly during holiday periods (2020 to 2021), when retail investors had time and attention. Conclusion: seasonality = short-term predictable patterns; a bubble = long-term overreaction; the two reinforce each other. Inflows that arrive at predictable calendar moments (bonuses in January, turn-of-the-month salaries, holidays) can add fuel to a bubble that is already running on narratives.

24. Lecture conclusion

Markets show "patterns" on two levels:

• Long-term: bubbles, narrative contagion, herding and overconfidence.
• Short-term: seasonality, driven by behavioral and institutional patterns.