## 1. Summary Performance Statistics for China Stock Markets, from Start of Market to

## Source details

**Canonical URL:** [1. Summary Performance Statistics for China Stock Markets, from Start of Market to](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0604.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0604.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0604.pdf.json)

---

### Institutional context and data
- China stock market characteristics:
  - Domestically traded local currency A stock market (accessible only to local investors until end-2002).
  - Hard-currency B stock market (accessible only to foreign investors until early-2001).
  - Different blocks of shares from the same issuer listed and traded in different markets (A and B) rather than dual listing of same shares.
- Market evolution:
  - Over the past decade the China stock market transformed to become the biggest stock market, by capitalization, in Asia outside Japan.
  - A-share index levels recorded sharp increases and A-share volatility reduced as the market developed; B-shares exhibited mediocre performance with initially low volatility increasing sharply during 2001.
- Data used:
  - About 12 years of daily index data from the start of the Shanghai and Shenzhen Stock Exchanges up to December 2002.
  - Sample periods:
    - Shanghai SSE A-shares: December 19, 1990 to December 2002.
    - Shenzhen SE A-shares: April 3, 1991 to December 2002.
    - Shanghai SE B-shares: February 21, 1992 to December 2002.
    - Shenzhen SE B-shares: March 3, 1992 to December 2002.
  - Indices are value-weighted; source: CSMAR.
- Institutional specifics:
  - Approximately 70 percent of shares held by government, SOEs, Chinese institutions and are non-tradable → low effective free-float.
  - Trading restrictions: daily maximum 10 percent price movement since 16 December 1996; prior price limit removal caused Shanghai A-index to jump 111 percent on 21 May 1992 (treated as outlier).
  - Settlement differences: A-shares t+1; B-shares t+3.
  - Market openness milestones: WTO accession December 2000; B market opened to domestic investors with hard currency February 2001; QFII December 2002.

### Research aims and methods
- Objectives:
  - Examine returns in both A and B stock markets for calendar anomalies (holiday, turn-of-the-year, day-of-the-week, half-year effects).
  - Determine impact of institutional and cultural differences on seasonality.
  - Assess role of less informed individual investors in observed inefficiencies.
  - Incorporate volatility, volume, and turnover to evaluate effects on pricing and liquidity.
- Return and other measures:
  - Close-to-close return: R_{t,C} = ln(I_{t,C}/I_{t-1,C}).
  - Open-to-close return: R_{t,O} = ln(I_{t,C}/I_{t,O}).
  - Volatility: conventional standard deviation of closing returns (equation (2)); other high-low based measures considered.
  - Liquidity: daily volume (Vol_t) and dollar turnover (Val_t); standardized abnormal volume AVol_t = (Vol_t - MAVol_t) / σ_{MAVol} computed over centered 90-day window excluding day of interest.
- Holiday and calendar definitions:
  - Half-month (turn-of-month): first-half = first 9 trading days; last-half = last 9 trading days; average trading days = 20.7 for Shanghai and Shenzhen.
  - Holiday window: three days prior and after each holiday; actual holiday date not included in pre- or post-holiday periods.
  - Main analysis confined to national holidays/observances; cultural festivals also categorized.
- Econometric approach:
  - Use non-parametric statistics and median tests due to non-normality and outliers.
  - De-seasonalization regressions to remove interacting effects: January residual regression (equation (5)), Half-month residual regression (equation (6)), Day-of-the-week residual regression (equation (7)), Holiday residual regression (equation (8)).
  - Year dummies Year_i (i = 1,...,11 for 1992–2002), month dummies Mnth_k (k = 1,...,11), DoW_n (n = 1,...,4). B-share regressions have 10 year dummies for markets starting in 1992.

### Key empirical findings
- Turn-of-the-year and monthly patterns:
  - Lack of a January effect in A markets and Shenzhen B market at 5 percent level.
  - Evidence of a February turn-of-the-year ("February effect"), consistent with Chinese Lunar New Year (CNY) timing.
  - All mean returns for February positive and large relative to medians and non-February means; Shanghai B February effect significant with mean and median returns significantly positive and corresponding significantly higher abnormal liquidity.
- Half-year effect:
  - Strong evidence of a "half-year effect" (February–June) in mean returns in all stock markets (A and B, Shanghai and Shenzhen).
  - Shenzhen A: median returns for first-half significantly positive.
  - Higher returns in A markets correspond with significantly higher liquidity; in B markets positive returns correspond with significantly higher risk.
- Half-month (turn-of-month) effect:
  - Some evidence in Shanghai A (median returns) but absent in most markets.
- Day-of-the-week effect:
  - Extremely strong in both A markets (best captured by median returns); less important in B markets (significant only at 10 percent).
  - A markets: biggest fall on Mondays; Fridays best; effect manifests over Friday-to-Monday nontrading period.
  - B markets: biggest declines on Tuesdays; Friday performance weaker; patterns suggest U.S. influence spilling into Tuesday.
  - All markets show lower volatility on Friday.
- Holiday effects and cultural distinctions:
  - Strong CNY holiday effect present and stronger compared with other public holidays, especially for A-shares.
  - Pre-holiday returns very significant and variances generally lower; liquidity significantly lower in A markets during pre-holiday period.
  - Post-holiday behavior differs: A markets sometimes positive post-holiday (returns remain positive post-CNY); B markets show conventional post-holiday negative returns.
  - Cultural festivals (DBF, MAF) generally show no observed holiday effect when exchanges remain open.
- Residual (adjusted) results:
  - Raw and residual results largely consistent; residual analyses confirm February, half-year, and day-of-the-week effects in A markets; B markets show significant February and holiday patterns especially in Shanghai B.
- Relative market behavior:
  - Cultural and structural factors both play important roles; A-shares and B-shares show different seasonal patterns consistent with investor composition and information access.

### Extension: Holiday effect (CNY and cultural holidays)
- Holiday categorizations:
  - Fixed state holidays (same date annually).
  - Non-fixed CNY holidays (lunar calendar).
  - All public holidays = fixed + CNY.
  - Cultural festivals (observed but not actual holidays; exchanges remain open).
- Findings:
  - Fixed state holidays: significant holiday effect for all markets except Shanghai A; effect stronger for B-shares.
  - CNY period: statistically significant holiday effect across all markets (mean and median returns) — higher returns not explained by higher volatility.
  - Liquidity pattern: lower liquidity pre-holiday and higher liquidity post-holiday consistent with positive holiday sentiment.
  - Post-holiday returns: A markets’ returns largely remain positive in post-holiday CNY period; B markets’ returns fall post-holiday (reversal).
  - Cultural festivals without exchange closure (DBF, MAF): generally no holiday effect.

### Investment strategies based on seasonalities (design and performance)
- Strategies considered (day-of-week excluded as non-tradeable):
  - S1: Invest in turn-of-the-year (February) effect.
  - S2: Invest in half-year (February to June) effect.
  - S3: Invest for one month from first day of CNY (positive returns persist into post-CNY period).
- Benchmarks:
  - B1: Invest in each non-February month for that year.
  - B2: Invest in July to November period.
  - B3: Invest in same time period in other months corresponding to CNY period.
- Random strategy comparisons:
  - R1: Random monthly investment = random start, hold one month (22 trading days); repeated 400 times.
  - R2: Random five-month investment = random start, hold five months (100 trading days); repeated 150 times.
  - R3: Random monthly investment repeated 400 times.
- Performance metrics:
  - Compounded rate of return over sample period, average buy-and-hold return per holding period, Sharpe ratio.
- Select empirical performance results:
  - February strategy (S1) compounded returns 1991–2002:
    - Shanghai A: almost 46 percent.
    - Shanghai B: 43 percent.
    - Shenzhen A: 42 percent.
    - Shenzhen B: 36 percent.
  - Half-year strategy (S2) compounded returns (Full 12-year period):
    - Shanghai A: 472 percent.
    - Shenzhen A: 412 percent.
    - Shanghai B: 466 percent.
    - Shenzhen B: 809 percent.
    - Benchmarks and randoms: benchmark B2 recorded a decline of 83 percent for Shanghai B; R2 returned 27 percent for Shanghai B; benchmark B2 almost down 70 percent for Shenzhen B; R2 returned 19 percent for Shenzhen B.
  - CNY month strategy (S3) compounded returns:
    - Shanghai A: 138 percent.
    - Shenzhen A: 87 percent.
    - Shanghai B: 50 percent.
    - Shenzhen B: 69 percent.
  - Per-month CNY averages:
    - Shanghai A: almost 9 percent per CNY month.
    - Shenzhen A: 5.9 percent per CNY month.
- Sharpe ratio ranking (Table 7; risk-adjusted performance):
  - Shanghai A, Half-Year Effect: 0.60
  - Shenzhen A, Holiday (CNY) Effect: 0.51
  - Shanghai A, February Effect: 0.50
  - Shenzhen B, Half-Year Effect: 0.47
  - Shenzhen A, Half-Year Effect: 0.46
  - Shanghai A, Holiday (CNY) Effect: 0.43
  - Shanghai B, Half-Year Effect: 0.42
  - Shenzhen A, February Effect: 0.38
  - Shanghai B, February Effect: 0.37
  - Shanghai B, Holiday (CNY) Effect: 0.32
  - Shenzhen B, Holiday (CNY) Effect: 0.31
  - Shenzhen B, February Effect: 0.27

### Risk, volatility, liquidity, and transaction costs
- Risk and volatility:
  - Standard deviations for selected strategies show substantially higher volatility for many high-return strategies, especially in B markets.
  - Higher returns in B markets partly explained by higher risk.
- Transaction costs and fees (Table 8, in basis points):
  - A stock markets: "round trip" trade transaction costs (including commissions to domestic brokers) estimated to reach as high as 200 basis points.
  - B stock markets: commissions estimated to range from 35 to 60 basis points, plus other fees:
    - Stamp: Shanghai 20.0, Shenzhen 20.0
    - Levy: Shanghai 26.0, Shenzhen 3.0
    - Administrative Fee: Shanghai 0.4, Shenzhen 5.0
    - Settlement Fee: Shanghai 5.0, Shenzhen 0.4
- Net-of-cost assessment:
  - Some strategies (notably S2 and S3 in selected markets) remain attractive after accounting for transaction costs (example: A-market round-trip up to 200 basis points relative to some five-month returns such as up to 22.3 percent for S2 in the Shenzhen A market).
  - High transaction costs may reduce attractiveness of lower-return strategies (for example, S1).

### Policy-relevant conclusions and research suggestions
- Main conclusions:
  - The lack of a January effect and presence of a February effect indicate CNY timing is central to turn-of-the-year seasonality in China.
  - Institutional segmentation (A vs. B) and cultural factors both influence seasonality; segmentation offers a natural experiment to distinguish cultural versus structural drivers.
  - A-share markets generally outperform B-share markets on a risk-adjusted basis across many seasonal strategies.
  - Structural spillovers (weekday effects tied to international markets) imply market-opening decisions affect transmission of external shocks.
- Policy and practitioner implications:
  - Models of returns and volatility for China markets should include calendar effects (CNY timing, half-year, day-of-the-week) and liquidity variables (volume, turnover).
  - Calendar-based trading strategies can generate economically significant returns in the A market even after transaction costs, suggesting exploitable inefficiencies tied to institutional structure and investor composition.
- Suggestions for future research:
  - Examine whether seasonalities converge as A stock markets are opened to foreign investors and B stock markets open to local investors with hard currency.
  - Study evolution of cultural effects as markets become more internationally integrated once sufficient post-opening data become available.

*Source: IMF working paper content (_wp0604) covering summary performance statistics and analysis of seasonalities in China stock markets through December 2002.*

### 1. Summary Performance Statistics for China Stock Markets, from Start of Market to

### _wp0604 - 1. Summary Performance Statistics for China Stock Markets, from Start of Market to

### Institutional context and data
- The China stock market is relatively new, less developed, and experienced rapid changes in its short history.
- Distinct institutional features:
  - Existence of a domestically traded local currency A stock market, which until the end of 2002 was accessible only to local investors.
  - Existence of a hard-currency B stock market, which until early 2001 was accessible only to foreign investors.
  - Different blocks of shares from the same issuer are listed and traded in the different stock markets (A and B) rather than a dual listing of the same shares.
- Market evolution:
  - Over the past decade the China stock market transformed to become the biggest stock market, by capitalization, in Asia outside Japan.
  - The domestic A stock market has experienced increased size, depth, and liquidity; A-share index levels recorded sharp increases and A-share volatility reduced as the market developed.
  - B-shares exhibited mediocre performance, with initially low volatility increasing sharply during the 2001 period; turnover and volume in the B market improved.
- Data used in the study:
  - About 12 years of daily index data from the start of the Shanghai and Shenzhen Stock Exchanges up to December 2002.
  - Subsample references in literature include January 1995 to December 1997 and start-of-trading to December 1993 in prior studies.

### Research aims and methods
- Objectives:
  - Examine returns in both A and B stock markets for evidence of calendar anomalies, including the holiday and turn-of-the-year effects.
  - Determine the impact of institutional and cultural differences (A-shares held solely by domestic investors; B-shares held by foreign investors) on seasonality.
  - Assess whether less informed individual investors contribute to observed inefficiencies.
  - Incorporate volatility, volume, and turnover to evaluate effects on pricing, liquidity, and seasonalities in volatility and volume.
- Methodological notes:
  - Tests include calendar-anomaly detection (holiday, turn-of-the-year, day-of-the-week, half-year effects) and evaluation of portfolio investment strategies accounting for transaction costs.
  - Volatility of returns used to capture uncertainty; volume and turnover used to determine liquidity/depth.

### Key empirical findings
- Turn-of-the-year effect:
  - Evidence points to a February—as opposed to a January—turn-of-the-year effect, partly because the Chinese Lunar New Year (CNY) normally occurs in late January or February.
- Holiday effects:
  - A strong CNY holiday effect is present.
  - CNY holiday effect appears stronger compared with other public holidays, especially for A-shares.
- Day-of-the-week effect:
  - Strong evidence for a day-of-the-week effect in China markets.
  - Internationally traded B stock markets tend to post negative returns on Tuesdays, corresponding to negative returns on Mondays in the United States.
  - The day-of-the-week effect in the “closed” A stock markets manifests over the Friday-to-Monday nontrading period.
- Half-year effect:
  - Strong evidence exists for a “half-year effect.”
- Relative market behavior:
  - Cultural and structural factors both play important roles in influencing pricing within the China market.
  - A-shares (domestic) and B-shares (international) show different seasonal patterns, consistent with differences in investor composition and information access.
- Investment strategy performance:
  - Investment strategies based on calendar anomalies indicate that A stock markets tend to be more profitable and offer economically significant returns.
  - These A-share profits persist even after accounting for transaction costs.

### Implications for market understanding and practice
- Cultural and institutional determinants:
  - Cultural festivals (CNY) materially affect return seasonality, highlighting the need to account for culturally specific holiday timing when analyzing seasonal effects.
  - Distinct investor bases (domestic vs. foreign) and information asymmetries likely drive differences in anomaly manifestations between A- and B-markets.
- Modeling and forecasting:
  - Models of returns and volatility for China markets should include calendar effects (CNY timing, half-year, day-of-the-week) and variables capturing liquidity (volume, turnover) and volatility.
- Trading and portfolio implications:
  - Calendar-based trading strategies can generate economically significant returns in the A market even after transaction costs, suggesting exploitable inefficiencies tied to institutional structure and investor composition.

*Source: IMF working paper content covering summary performance statistics and analysis of seasonalities in China stock markets through December 2002.*

### Section VI. In Section VII, we use our findings to derive and test possible investment

### _wp0604 - Section VI. In Section VII, we use our findings to derive and test possible investment

### II. LITERATURE ON SEASONALITIES
- Definition: Seasonalities or calendar anomalies are significantly different share market returns at distinct cusps in time (days of week, periods of month, or year).
- January (Turn-of-the-Year) Effect:
  - Observed decline in December followed by increase in January; documented by Branch (1977), Dyl (1977), Keim (1983), Reinganum (1983), Roll (1983).
  - Proposed explanations: (i) tax-loss selling; (ii) small-firm premium concentrated in January.
  - Interaction with other anomalies: half of small-firm premium in January occurs in first few days (Keim, 1983); day-of-the-week effect may occur primarily during January (Rogalski, 1984b).
  - Hypothesis of relocation/moderation: January effect may have moved into November and December due to mutual fund holding reports at end of October (Bernstein, 1999).
  - Empirical generality: January historically best month across many markets; some countries have January returns greater than annual average (Agrawal and Tandon, 1994; Gultekin and Gultekin, 1983).
- Half-Month (Turn-of-the-Month) Effect:
  - Higher returns in first-half relative to last-half of trading month (Ariel, 1987; Agrawal and Tandon, 1994).
  - Hensel and Ziemba (1996): higher returns on last day and first four days of month; from 1928–1993 total S&P 500 return mostly during turn of month.
  - Proposed mechanism: bulk of cash flows at end of month (salaries, interest), benefiting investors who invest prior to turn of month.
- Day-of-the-Week (Weekend) Effect:
  - Unequal daily mean returns; Monday often the worst, Friday often best (Fields, 1931; Cross, 1973; French, 1980; Gibbons and Hess, 1981; Keim and Stambaugh, 1984).
  - Intraday pattern: first 45 minutes prices fall on Monday, rise on other days (Harris, 1986).
  - International evidence: lowest mean returns may be Tuesday in Australia/Japan (Jaffe and Westerfield, 1985); generally lower Tuesday and higher Friday in Europe; Asia-Pacific shows lower Tuesday and higher Friday.
  - Explanations: (i) settlement effect; (ii) information release; (iii) measurement error; (iv) trading behavior; (v) interaction with other calendar anomalies; (vi) moods of participants.
  - Trading costs often make exploitation virtually impossible (Chen, Kwok and Rui, 2001; Jacobs and Levy, 1988).
- Holiday Effect:
  - Higher-than-normal returns, mainly pre-holiday (Lakonishok and Smidt, 1988; Ariel, 1990).
  - Examples: roughly half of annual DJIA gain occurs in 10 pre-holiday trading days; over one-third of US positive returns each year in eight trading days prior to market-closed holidays.
  - Explanations: (i) holiday effect may capture other seasonalities; (ii) positive holiday sentiment reducing trading/closing out positions (Fabozzi, Ma and Briley, 1994).
  - Cultural context: stronger holiday effect around cultural holidays in Malaysia, Singapore, India, Thailand (Chan, Khanthavit and Thomas, 1996); cultural holidays (e.g., CNY) linked to abnormal returns in Hong Kong, Japan, Malaysia, Singapore, Korea, Taiwan Province of China (Cadsby and Ratner, 1992; Yen and Shyy, 1993).

### III. INSTITUTIONAL ASPECTS OF CHINESE STOCK MARKET
- Two main institutional aspects:
  - Closed nature of A-share markets (historical restrictions on foreign investment; changes: WTO accession December 2000; B market opened to domestic investors with hard currency February 2001; QFII December 2002).
  - Segmentation between A- and B-shares.
- Market structure and implications:
  - Approximately 70 percent of shares held by government, SOEs, Chinese institutions and are non-tradable → low effective free-float and artificially strong domestic demand.
  - Trading restrictions: daily maximum 10 percent price movement since 16 December 1996; prior price limit removal caused Shanghai A-index to jump 111 percent on 21 May 1992 (treated as outlier).
  - Segmentation effects:
    - Settlement: A-shares t+1; B-shares t+3 → B-share returns on Friday should be higher to compensate for settlement delay.
    - B-share foreign investors have more timely information and B markets are smaller, less liquid, more volatile.
  - Insularity: A markets largely insulated from short-term U.S. market influence; B markets more exposed.

### IV. DATA AND RESEARCH METHOD
- Data:
  - Source: China Stock Market and Accounting Research (CSMAR) database.
  - Sample periods:
    - Shanghai SSE A-shares: December 19, 1990 to December 2002.
    - Shenzhen SE A-shares: April 3, 1991 to December 2002.
    - Shanghai SE B-shares: February 21, 1992 to December 2002.
    - Shenzhen SE B-shares: March 3, 1992 to December 2002.
  - Indices are value-weighted.
  - Data screening: missing values checked; zero returns verified; outliers (e.g., 21 May 1992 event) excluded.
- Variables analyzed: share price return, volatility, market liquidity.
- Return measures:
  - Close-to-close return: R_{t,C} = ln(I_{t,C}/I_{t-1,C}) (equation (1)).
  - Open-to-close return: R_{t,O} = ln(I_{t,C}/I_{t,O}) (equation (4)).
  - Equally-weighted vs value-weighted returns calculated.
- Definitions and classifications:
  - Half-month (turn-of-month): trading month = last day previous calendar month (inclusive) to last day current calendar month (exclusive); first-half = first 9 trading days; last-half = last 9 trading days; average trading days = 20.7 for Shanghai and Shenzhen.
  - Holiday window: three days prior and after each holiday; actual holiday date not included in pre- or post-holiday periods.
  - Main analysis confined to national holidays/observances.
- Volatility and liquidity measures:
  - Volatility: conventional standard deviation of closing returns reported (equation (2)); other high-low based measures considered but largely consistent.
  - Liquidity: daily volume (Vol_t) and dollar turnover (Val_t).
  - Time trends in volume/turnover allowed for; 90-day moving average used to capture medium-term dependence.
  - Standardized abnormal volume AVol_t = (Vol_t - MAVol_t) / σ_{MAVol} with MAVol and σ computed over centered 90-day window excluding day of interest (equation (3)). AVal_t computed analogously using dollar turnover.
- Measurement/econometric issues addressed:
  - Use non-parametric statistics due to non-normality and large sample sizes.
  - Median tests to mitigate outlier influence.
  - Recompute open-to-close returns to isolate overnight/weekend effects.
  - De-seasonalization regressions to remove interacting effects:
    - January residual regression (equation (5)).
    - Half-month residual regression (equation (6)).
    - Day-of-the-week residual regression (equation (7)).
    - Holiday residual regression (equation (8)).
  - Year dummies Year_i (i = 1,...,11 for 1992–2002), month dummies Mnth_k (k = 1,...,11), DoW_n (n = 1,...,4). B-share regressions have 10 year dummies for markets starting in 1992.

### V. RESULTS
- Visual patterns (Figure 4):
  - Little visual January effect for Shenzhen A and both B markets; higher returns often in February → possible “February effect”.
  - Mean daily returns generally higher in first-half of year (February to June) across markets → “half-year effect”.
  - Shanghai A mean returns largely positive across months.
- Unadjusted (raw) returns (Table 3, five panels):
  - January effect:
    - No January effect at 5 percent level in either A market or Shenzhen B market.
    - Shanghai B: significantly lower mean returns in January and significantly higher risk levels.
    - A markets: abnormal liquidity significantly lower during January.
  - February effect:
    - All mean returns for February positive and large relative to medians and non-February means, but not statistically significant for A markets and Shenzhen B.
    - Shanghai B: February effect significant with mean and median returns significantly positive; corresponds to significantly higher abnormal liquidity; not explained by higher risk.
  - Half-year effect:
    - Evident in mean returns in all stock markets (A and B, Shanghai and Shenzhen).
    - Shenzhen A: median returns for first-half significantly positive.
    - Higher returns in A markets correspond with significantly higher liquidity; B markets’ positive returns correspond with significantly higher risk.
  - Half-month effect:
    - Some evidence in Shanghai A (median returns) but not in others.
  - Day-of-the-week effect:
    - Extremely strong in both A markets (best captured by median returns); less important in B markets (significant only at 10 percent).
    - A markets: biggest fall on Mondays; Fridays best.
    - B markets: biggest declines on Tuesdays; Friday performance weaker relative to A markets.
    - All markets show lower volatility on Friday.
    - Interpretation: B-share patterns suggest U.S. influence; A-share similarity indicates domestic patterns despite segmentation.
  - Holiday effect:
    - Very significant positive pre-holiday returns, especially in B and Shenzhen A markets; Shanghai A pre-vs-post returns insignificant.
    - Pre-holiday variances generally lower; liquidity significantly lower in A markets during pre-holiday period.
    - Post-holiday behavior differs: A markets sometimes positive post-holiday; B markets show conventional post-holiday negative returns.
- Adjusted (residual) returns (Table 4):
  - Little interaction dependence; raw and residual results largely consistent.
  - January effect: little evidence across markets; Shanghai B significantly negative in January and significantly positive in February → February effect.
  - February residual returns: substantially positive across markets though not always statistically significant; higher liquidity during February; not explained by higher volatility.
  - Half-year residual effect: significant in all markets except Shanghai A (median and mean returns).
  - Half-month residual effect: no evidence.
  - Day-of-the-week residual effect: very significant for adjusted mean returns in A markets (Fridays positive, Mondays most negative); B markets show weaker day-to-day differences with consistent negative adjusted mean returns on Tuesdays.
  - Holiday residual effect: very significant for all markets except Shanghai A; positive pre-holiday returns reverse in B markets post-holiday and in Shenzhen A, but not clearly in Shanghai A.

### VI. EXTENSION: HOLIDAY EFFECT
- Focus: deeper examination of Chinese Lunar New Year (CNY) and other cultural festivals (Mid-Autumn Festival (MAF), Dragon Boat Festival (DBF)); Clear Brightness Festival (CBF) and Hungry Ghost Festival (HGF) excluded as inauspicious.
- Holiday categorizations (Table 5):
  - Fixed state holidays (same date annually).
  - Non-fixed CNY holidays (lunar calendar).
  - All public holidays (fixed + CNY).
  - Cultural festivals (observed but not actual holidays; exchanges remain open).
- Findings:
  - Fixed state holidays: significant holiday effect for all markets except Shanghai A; effect stronger for B-shares.
  - CNY period: statistically significant holiday effect across all markets (mean and median returns) — higher returns not explained by higher volatility.
  - Liquidity: evidence of lower liquidity pre-holiday and higher liquidity post-holiday consistent with positive holiday sentiment.
  - Post-holiday behavior:
    - A markets: returns largely remain positive in post-holiday CNY period.
    - B markets: returns fall in post-holiday CNY period (conventional reversal).
  - Cultural festivals (DBF, MAF): generally no observed holiday effect — consistent with prior findings that pre-holiday abnormal returns require exchange closure.
  - Conclusion: holiday effect in B shares across all holidays consistent with international markets; in segmented A markets anomaly significant only for cultural state holiday CNY.

### VII. FURTHER EXTENSIONS: INVESTMENT STRATEGIES BASED ON SEASONALITIES
- Considered investment strategies based on observed seasonalities (day-of-week excluded as non-tradeable):
  - S1: Invest in turn-of-the-year (February) effect.
  - S2: Invest in half-year (February to June) effect.
  - S3: Invest for one month from first day of CNY (positive returns persist into post-CNY period).
- Benchmarks:
  - B1: Invest in each non-February month for that year.
  - B2: Invest in July to November period.
  - B3: Invest in same time period in other months corresponding to CNY period.
- Random strategy comparisons:
  - R1: Random monthly investment = random start, hold one month (22 trading days); repeated 400 times.
  - R2: Random five-month investment = random start, hold five months (100 trading days); repeated 150 times.
  - R3: Random monthly investment (same as R1) repeated 400 times.
- Performance metrics computed:
  - Compounded rate of return for each strategy over sample period.
  - Average buy-and-hold return per holding period.
  - Sharpe ratio for each strategy.
- Key empirical result (Panel A of Table 6, February effect):
  - Compounded returns for S1 (1991–2002):
    - Shanghai A: almost 46 percent.
    - Shanghai B: 43 percent.
    - Shenzhen A: 42 percent.
    - Shenzhen B: 36 percent.
  - Strategy comparisons: S1 compares well with benchmarks B1 and random R1 across markets; B1 posted a negative return for Shanghai B over the 12-year period (additional comparative details appear in Table 6 but are not reproduced here).

*Italicized source attribution: _wp0604 - Section VI. In Section VII, we use our findings to derive and test possible investment*

### 0.4 percent for the Shenzhen B market.

### _wp0604 - 0.4 percent for the Shenzhen B market.

### Key empirical findings on seasonalities and market structure
- The China stock markets show both cultural and structural influences on stock-pricing behavior.
- The lack of a January effect and the presence of a February effect suggests the “turn-of-the-year” may be around the Chinese Lunar New Year (CNY), which usually begins in late January or sometime during February.
- A significant half-year effect is observed across all markets for the five months beginning in February (February–June).
- The half-month (turn-of-the-month) effect is absent in China stock markets, consistent with cultural calendar differences (Chinese calendar centered on lunar phases).
- The day-of-the-week effect shows structural influences:
  - A-shares (domestic-invested) show the effect over Friday-to-Monday.
  - B-shares (internationally traded) tend to reflect overnight U.S. developments—negative returns in the U.S. on Monday spill over into Tuesday returns in B markets.
- The holiday effect differs between A- and B-shares:
  - A-shares: strong holiday (CNY) effect around cultural state holidays; post-CNY returns tend to be positive.
  - B-shares: holiday effect observed around both cultural and non-cultural public holidays; post-holiday negative-return behavior is consistent with other international markets.

### Investment strategy performance (select results from Tables 6 and 7)
- February (monthly) strategy (S1):
  - Strategy S1 produced average monthly returns in excess of 3 percent for Shanghai B shares (economically significant).
  - An investment in the Shanghai A stock market in each February from 1991 to 2002 would have returned a compounded rate of almost 46 percent over the period.
- Half-year (5-month) strategy (S2, February–June):
  - Shanghai A: compounded return 472 percent (Full 12-year period).
  - Shenzhen A: compounded return 412 percent.
  - Shanghai B: compounded return 466 percent; benchmark B2 recorded a decline of 83 percent; R2 returned 27 percent.
  - Shenzhen B: compounded return 809 percent; benchmark B2 was down almost 70 percent; R2 returned 19 percent.
  - Despite exceptionally high compounded returns in some B markets (e.g., Shenzhen B 809 percent), higher risks partly explain these outcomes.
- Chinese New Year (CNY) month strategy (S3):
  - Shanghai A: compounded return 138 percent (CNY month each year).
  - Shenzhen A: compounded return 87 percent.
  - Shanghai B: compounded return 50 percent.
  - Shenzhen B: compounded return 69 percent.
  - On a per-month basis during CNY, Shanghai A returned almost 9 percent and Shenzhen A returned 5.9 percent (average per CNY month).
- Sharpe ratio ranking (Table 7) — relative risk-adjusted performance:
  - Shanghai A, Half-Year Effect: 0.60
  - Shenzhen A, Holiday (CNY) Effect: 0.51
  - Shanghai A, February Effect: 0.50
  - Shenzhen B, Half-Year Effect: 0.47
  - Shenzhen A, Half-Year Effect: 0.46
  - Shanghai A, Holiday (CNY) Effect: 0.43
  - Shanghai B, Half-Year Effect: 0.42
  - Shenzhen A, February Effect: 0.38
  - Shanghai B, February Effect: 0.37
  - Shanghai B, Holiday (CNY) Effect: 0.32
  - Shenzhen B, Holiday (CNY) Effect: 0.31
  - Shenzhen B, February Effect: 0.27

### Risk, volatility, liquidity, and transaction costs
- Standard deviations for selected strategies and markets show substantially higher volatility for many high-return strategies, especially in B markets (e.g., standard deviation figures reported in Table 6 panels).
- Transaction costs:
  - A stock markets: transaction costs for a “round trip” trade (including commissions to domestic brokers) are estimated to reach as high as 200 basis points.
  - B stock markets: commissions estimated to range from 35 to 60 basis points, on top of other costs shown in Table 8.
  - Table 8 (Transaction Costs for China B Stock Markets, in basis points):
    - Stamp: Shanghai 20.0, Shenzhen 20.0
    - Levy: Shanghai 26.0, Shenzhen 3.0
    - Administrative Fee: Shanghai 0.4, Shenzhen 5.0
    - Settlement Fee: Shanghai 5.0, Shenzhen 0.4
- Net-of-cost assessment:
  - Some strategies (notably S2 and S3 in selected markets) remain attractive even after accounting for transaction costs (A-market round-trip up to 200 basis points pales relative to some five-month returns such as up to 22.3 percent for S2 in the Shenzhen A market as shown in Table 6).
  - High transaction costs may reduce the attractiveness of strategies with lower returns (for example, S1).

### Policy-relevant conclusions and research suggestions
- Institutional segmentation and cultural factors both influence seasonality in China stock markets; segmentation (A vs. B) provides a natural experiment to distinguish cultural versus structural drivers.
- Key conclusions for practitioners and policymakers:
  - A-share markets generally outperform B-share markets on a risk-adjusted basis across many seasonal strategies.
  - The Chinese Lunar New Year (CNY) timing is central to turn-of-the-year seasonality in China, rather than the calendar January effect observed elsewhere.
  - The presence of structural spillovers (weekday effects tied to international markets) suggests policy and market-opening decisions can affect transmission of external shocks.
- Suggestions for future research:
  - Examine whether seasonalities converge as A stock markets are opened to foreign investors and B stock markets open to local investors with hard currency.
  - Study how cultural effects evolve as markets become more internationally integrated once sufficient post-opening data become available.

*Source: _wp0604 - 0.4 percent for the Shenzhen B market.*

### 1.  Fixed State Holidays×√×√Structural

### 1.  Fixed State Holidays×√×√Structural

### Holiday Classification
- Fixed State Holidays: ×√×√Structural
- Non-Fixed CNY Holidays: √√√√Cultural
- Non-Holiday Cultural Festivals: ××××None

### Stock Market — Indices (Figure 1: Shanghai and Shenzhen Stock Market Indices from Start of Market to December 2002)
- A-Shares scales and labels:
  - Vertical scale left (Shanghai Index): 0, 500, 1,000, 1,500, 2,000, 2,500
  - Vertical scale right (Shenzhen Index): 0, 100, 200, 300, 400, 500, 600, 700, 800
  - Time axis ticks: Dec-90, Dec-91, Dec-92, Dec-93, Dec-94, Dec-95, Dec-96, Dec-97, Dec-98, Dec-99, Dec-00, Dec-01, Dec-02
  - Series shown: SHSE A-Share (LHS), SZSE A-Share (RHS)
- B-Shares scales and labels:
  - Vertical scale left (Shanghai Index): 0, 50, 100, 150, 200, 250, 300
  - Vertical scale right (Shenzhen Index): 0, 50, 100, 150, 200, 250, 300, 350, 400, 450, 500
  - Time axis ticks: Dec-91, Dec-92, Dec-93, Dec-94, Dec-95, Dec-96, Dec-97, Dec-98, Dec-99, Dec-00, Dec-01, Dec-02
  - Series shown: SHSE B-Share (LHS), SZSE B-share (RHS)
- Source: CSMAR

### Stock Market — Daily Returns (Figure 2: Shanghai and Shenzhen Stock Markets: Daily Returns from Start of Market to December 2002; In percent)
- A-Shares returns scales and labels:
  - Vertical scales for returns: -20, -10, 0, 10, 20, 30, 40
  - Alternative scale for Shenzhen Returns (%): -60, -50, -40, -30, -20, -10, 0, 10, 20, 30, 40
  - Time axis ticks: Dec-90, Dec-91, Dec-92, Dec-93, Dec-94, Dec-95, Dec-96, Dec-97, Dec-98, Dec-99, Dec-00, Dec-01, Dec-02
  - Series shown: SHSE A-Share Returns (LHS), SZSE A-Share Returns (RHS)
  - Moving averages: 90 Day MA (SHSE A-Share), 90 Day MA (SZSE A-Share)
- B-Shares returns scales and labels:
  - Vertical scales: -15, -10, -5, 0, 5, 10, 15, 20
  - Alternative scale for Shenzhen Returns (%): -50, -40, -30, -20, -10, 0, 10, 20
  - Time axis ticks: Dec-91, Dec-92, Dec-93, Dec-94, Dec-95, Dec-96, Dec-97, Dec-98, Dec-99, Dec-00, Dec-01, Dec-02
  - Series shown: SHSE B-Share Returns (LHS), SZSE B-Share Returns (RHS)
  - Moving averages: 90 Day MA (SHSE B-Share), 90 Day MA (SZSE B-Share)
- Source: CSMAR

### Stock Market — Volume and Turnover (Figure 3: Shanghai and Shenzhen Stock Markets: Daily Volume and Turnover from Start of Market to December 2002)
- A-Shares volume and turnover scales and labels:
  - Volume (Millions of shares) scale: 0, 500, 1,000, 1,500, 2,000, 2,500, 3,000, 3,500, 4,000, 4,500, 5,000
  - Turnover Value Traded (Millions of U.S. dollars) scale: -3,000, -2,000, -1,000, 0, 1,000, 2,000, 3,000, 4,000, 5,000, 6,000, 7,000
  - Time axis ticks: Dec-90, Dec-91, Dec-92, Dec-93, Dec-94, Dec-95, Dec-96, Dec-97, Dec-98, Dec-99, Dec-00, Dec-01, Dec-02
  - Series shown: SHSE A-Share - Volume, SZSE A-Share - Volume, SHSE A-Share - USD VALUES, ZSE A-Share - USD Value
  - Additional lines: Polynomial Trend (SHSE A-Share - Volume), 90 Day MA (SHSE A-Share - USD Value)
- B-Shares volume and turnover scales and labels:
  - Volume (Millions of shares) scale: 0, 200, 400, 600, 800, 1,000, 1,200, 1,400
  - Turnover Value Traded (Millions of U.S. dollars) scale: -400, 100, 600, 1,100, 1,600, 2,100
  - Time axis ticks: Dec-91, Dec-92, Dec-93, Dec-94, Dec-95, Dec-96, Dec-97, Dec-98, Dec-99, Dec-00, Dec-01, Dec-02
  - Series shown: SHSE B-Share - Volume, SZSE B-Share - Volume, SHSE B-Share - USD Value, SZSE B-Share - USD Value
  - Additional lines: Polynomial Trend (SHSE B-Share - Volume), 90 Day MA (SHSE B-Share - USD Value)
- Source: CSMAR

### Stock Market — Seasonal Pattern (Figure 4: China Stock Markets: Mean Daily Close-to-Close Returns by Month, January 1993 to December 2002; In percent)
- Months: Jan, Feb, Mar, April, May, June, July, Aug, Sep, Oct, Nov, Dec
- Series shown: SHSE A-shares, SZSE A-Shares, SHSE B-shares, SZSE B-Shares
- Daily Returns Close-to-Close (%) scale: -0.50, -0.40, -0.30, -0.20, -0.10, 0.00, 0.10, 0.20, 0.30, 0.40
- Time period covered: January 1993 to December 2002

*Source: _wp0604 - 1.  Fixed State Holidays×√×√Structural (figures and tables sourced to CSMAR).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0604.pdf_