Earnings Surprise Drift in Taiwan: Big Misses, Big Drift

Taiwan's Q5-Q1 earnings drift spread of +6.25% at T+63 is the second-largest globally in our study vs the TAIEX. The spread has contributions from both sides. Biggest misses produce -3.53% CAR at T+63 (t=-14.4). Biggest beats produce +2.72% (t=8.4). Taiwan's PEAD is miss-heavy but both tails are statistically significant.

Contents

1. Method
2. The Strategy
3. What We Found
4. Positive vs Negative Surprise Drift
5. Quintile Analysis at T+63
6. The Asymmetry in Context
7. When It Works and When It Struggles
8. Run It Yourself
9. Limitations
10. Part of a Series
11. References

Data: FMP financial data warehouse, 2000–2025. Updated March 2026.

Method

The Strategy

Post-Earnings Announcement Drift (PEAD) documents that stocks continue drifting in the direction of their earnings surprise for weeks after announcement. Ball and Brown (1968) first established this pattern. Bernard and Thomas (1989) confirmed the drift persists for up to 60 trading days. Foster, Olsen, and Shevlin (1984) analyzed different earnings windows.

Taiwan's market has a heavy technology concentration, dominated by semiconductor and electronics manufacturers. Analyst coverage is meaningful, with a significant domestic sell-side community. The market's retail participation is high relative to most developed markets, which can affect how quickly information is priced in.

What We Found

Positive vs Negative Surprise Drift

Taiwan's beat rate is 43.7%: more companies miss than beat vs the TAIEX. The negative surprise drift is the larger signal. Misses produce -2.03% at T+63. Positive surprises produce +1.54%. The positive drift builds steadily through all windows (+0.56% at T+1, +0.88% at T+5, +1.08% at T+21, +1.54% at T+63).

The negative drift shows its sharpest move early (-1.01% by T+5), stabilizes briefly, then accelerates to -2.03% by T+63. This is consistent with a market that partially prices in the miss immediately, stabilizes briefly, then resumes drifting as institutional investors reduce positions over the following months.

Quintile Analysis at T+63

The Q5-Q1 spread is +6.25% at T+63, the second-largest globally after India (+6.89%). Q1's t-stat of -14.4 is among the strongest in our global study. Q5 (t=8.4) is also statistically robust.

The Q5 hit rate of 52.3% is above 50%. The positive Q5 average (+2.72%) is carried by a majority of positive outcomes. This contrasts with Japan's Q5 hit rate of 56.6%, where the majority of big-beat events produce positive drift.

Q3 (near-consensus surprises) produces -1.17% at T+63, which is unusual. This further confirms a general downward bias in the Taiwan sample, likely reflecting a market where missing estimates is the norm and even "fair" results can disappoint.

The Asymmetry in Context

Taiwan's miss-side leads but both sides contribute. The Q1 miss magnitude (-3.53%) is larger than Japan's Q1 (-2.84%) and India's Q1 (-1.50%). The Q5 beat (+2.72%) exceeds Japan's Q5 (+2.09%) but trails India's Q5 (+5.39%).

When It Works and When It Struggles

Taiwan's technology concentration means earnings surprises often carry supply chain information. A semiconductor company missing estimates isn't just company-specific news. It can signal industry-wide demand weakness that takes time to propagate through analyst forecasts. This may explain the late-accelerating negative drift pattern.

The beat side is meaningful. Q5 beats (+2.72%) are statistically significant with a 52.3% hit rate. Long-only PEAD strategies in Taiwan are viable on the beat side.

Q1 miss avoidance (or shorting) is the cleaner signal. The Q1 t-stat of -11.4 is very high, the drift is strong (-3.71%), and the hit rate (33.0%) confirms most worst-miss events in Taiwan produce sustained negative drift.

Run It Yourself

python3 earnings-surprise/screen.py --preset taiwan

-- Recent earnings surprises on Taiwan exchanges
WITH deduped AS (
  SELECT
    es.symbol,
    es.date,
    es.epsActual,
    es.epsEstimated,
    (es.epsActual - es.epsEstimated) / NULLIF(ABS(es.epsEstimated), 0) AS std_surprise,
    ROW_NUMBER() OVER (PARTITION BY es.symbol, es.date ORDER BY es.lastUpdated DESC) AS rn
  FROM earnings_surprises es
  JOIN profile p ON es.symbol = p.symbol
  WHERE p.exchange IN ('TAI', 'TWO')
    AND p.isActivelyTrading = true
    AND es.date >= CURRENT_DATE - INTERVAL '90 days'
    AND ABS(es.epsEstimated) > 0.01
)
SELECT
  symbol,
  date,
  epsActual,
  epsEstimated,
  ROUND(std_surprise * 100, 1) AS surprise_pct
FROM deduped
WHERE rn = 1
  AND ABS(std_surprise) > 0.10
ORDER BY date DESC, std_surprise ASC
LIMIT 50

Run this query on Ceta Research →

Limitations

17,214 events provides solid statistical power, but the miss-heavy event distribution (9,393 negative vs 7,303 positive) means the quintiles aren't perfectly balanced. Q5 has 2,998 events vs Q1's 3,042.

Taiwan's heavy technology sector weighting means the results may be partially capturing sector dynamics rather than pure earnings surprise drift. A sector-controlled version of this analysis might show different quintile patterns.

The Q5 hit rate of 46.4% is worth emphasizing. Practitioners running a long-only Q5 strategy in Taiwan face a below-coin-flip probability of positive outcomes at the individual event level. The positive average is real but driven by tail events. Position sizing matters more here than in markets with stronger hit rates.

Part of a Series

• PEAD Across 14 Global Exchanges: The Drift Is Universal
• Earnings Surprise Drift on US Stocks: 170,000 Events, 26 Years

References

• Ball, R. & Brown, P. (1968). An empirical evaluation of accounting income numbers. Journal of Accounting Research.
• Bernard, V. & Thomas, J. (1989). Post-earnings-announcement drift: Delayed price response or risk premium? Journal of Accounting Research.
• Foster, G., Olsen, C. & Shevlin, T. (1984). Earnings releases, anomalies, and the behavior of security returns. The Accounting Review.

Data: Ceta Research, FMP financial data warehouse