Korean equity dividends follow a seasonal pattern with no close parallel in US or European markets. The overwhelming majority of KOSPI-listed companies declare and distribute a single annual dividend tied to the December 31 fiscal year-end, with the ex-dividend date (배당락일) falling on the last or second-to-last trading day of December. For any rotation strategy that touches KOSPI constituents between mid-December and mid-January, this clustering creates a systematic pricing effect that a naive back-test will either ignore entirely or handle incorrectly — and either way, the P&L is wrong.
This post covers how the KRX dividend calendar is structured, where back-test implementations most commonly fail to align ex-dates to KRX settlement mechanics, and what the practical magnitude of the error looks like on a representative strategy.
Past-performance disclaimer: Back-test illustrations in this article use synthetic parameter sets. Backtest results are not a guarantee of future returns; this is research, not investment advice.
How the KRX Dividend Calendar Is Actually Structured
KRX operates on a T+2 settlement cycle (결제일). To be entitled to a dividend, an investor must hold shares at the close of trading two business days before the record date (배당기준일). For Korean companies with a December 31 record date — which covers the majority of KOSPI 200 constituents — the ex-dividend date falls on December 28 or the last trading day of December minus two business days, depending on the year's trading calendar.
The clustering effect is significant. In a typical year, 60–70% of KOSPI 200 constituents by count go ex-dividend within a two-week window straddling late December. The price adjustment mechanism on ex-date is mechanical: the KRX opening reference price (기준가격) is reduced by the per-share dividend amount on the morning following the ex-date. For stocks with a 1.5–3% dividend yield, this creates a visible step-down in the price series on a single calendar day.
Where back-testers typically go wrong is in confusing the announced dividend date with the adjusted price date. Most publicly available OHLCV data feeds for Korean equities provide dividend-adjusted close prices, but the adjustment is applied to the historical series at the time of payout — not at the time of announcement. For a back-test consuming daily close prices, the adjusted close on December 27 and the adjusted close on December 28 (ex-date) should show the dividend step-down. If the back-test is using unadjusted prices, the December 28 close will show an apparent drop of 2% that the strategy logic may misinterpret as a price momentum signal pointing short.
The Settlement-Lag Problem in Rotation Signal Timing
Consider a monthly momentum rotation that ranks KOSPI 200 constituents by 12-1 month return and rebalances on the first trading day of each month. For a January 2 rebalance (after December 31 settlement), the momentum scores computed on December 31 closing prices should be based on dividend-adjusted prices — otherwise, the ranking artificially penalizes high-dividend names that went ex in late December.
The subtler issue is T+2 settlement timing. If an investor sells a KOSPI position on December 26 (before ex-date), settlement completes December 30. The investor does not receive the dividend. If they sell on December 27 — the day before ex-date — settlement completes December 31, and they hold through ex-date and receive the dividend. This one-day difference changes the effective return for that holding period by the full dividend yield. A back-test that applies a uniform "sell on signal date, receive dividend if held through year-end" rule without verifying KRX settlement dates will mis-attribute dividend income to a subset of simulated trades that, in real execution, would not have qualified.
import pandas as pd
# Load KRX trading calendar (available from KRX Data Service)
krx_calendar = pd.read_csv('krx_trading_days.csv', parse_dates=['date'])
trading_days = krx_calendar['date'].tolist()
def ex_date_for_fiscal_year_end(record_date: pd.Timestamp) -> pd.Timestamp:
"""
Returns the KRX ex-dividend date given a record date.
Ex-date = record_date minus 2 KRX business days (T+2 settlement).
"""
record_idx = trading_days.index(record_date)
return trading_days[record_idx - 2]
# Example: FY2023 record date Dec 31, 2023
record = pd.Timestamp('2023-12-29') # last trading day of Dec 2023
ex_date = ex_date_for_fiscal_year_end(record)
print(f"Ex-dividend date: {ex_date.date()}") # Expected: 2023-12-27Interim Dividends and the Mid-Year Edge Cases
While December clustering is the dominant pattern, a growing minority of KOSPI 200 companies — particularly financials (금융주) and large-cap consumer names — pay interim dividends (중간배당) in June or July aligned to a June 30 interim record date. These create a smaller but structurally identical T+2 alignment problem in late June, which matters for any strategy rebalancing around the mid-year period.
Additionally, physical spin-offs (물적분할) and corporate splits (인적분할) create what look like dividend-like price step-downs in the daily series, but are definitionally not dividends and should not be dividend-adjusted the same way. A back-test that naively applies dividend adjustment logic to all price step-downs will incorrectly treat a 인적분할 as income-generating, inflating historical P&L for any strategy that held the parent company through the split date. KRX publishes corporate action events separately through its Data Service, and a reliable back-test framework must distinguish between cash dividend adjustments and spin-off adjustments at the data layer.
Magnitude: How Much Does Misalignment Actually Cost?
We are not claiming that dividend misalignment is the dominant source of back-test error — we are saying it is a systematic bias that compounds predictably with strategy turnover and KOSPI sector exposure. The effect is largest for strategies that:
- Hold high-dividend-yield names (banks, utilities, large-cap industrials) through December
- Rebalance in January using prices computed from December 31 data
- Have high turnover, because each rebalance through December creates another ex-date alignment decision
For a synthetic illustration: a KOSPI 200 momentum strategy with quarterly rebalance and 40% annual turnover, running 2010–2023 on unadjusted prices versus KRX-calendar-aligned dividend-adjusted prices, shows a P&L difference in the range of 60–120 basis points in years with high KOSPI dividend payout ratios (typically 2015–2018 and 2021–2022 when large-cap payouts were elevated). In years where the ex-date falls on the very last trading day of December, the alignment error is highest because the strategy has the least margin for T+2 settlement correction.
What a Correctly Aligned Implementation Does
A proper implementation maintains three separate price series: raw unadjusted OHLCV, dividend-adjusted OHLCV for signal computation, and total-return series (dividend-adjusted plus cash dividend reinvested) for P&L calculation. The distinction matters because signal computation and P&L attribution use different series for coherent reasons — momentum ranks should use price-only adjusted series so that a stock's 12-month return rank is not inflated by a large one-time special dividend; P&L should use total return to fairly credit dividend income to holding periods that earned it.
Finology's data layer separates these series and aligns all dividend events to the KRX settlement calendar, including T+2 ex-date calculation, interim dividend tagging, and corporate action flagging (물적분할 / 인적분할 excluded from dividend adjustment). The December ex-date clustering is handled at the data ingestion stage rather than patched at the strategy layer, which means the error cannot silently re-emerge when a new strategy is tested.
The KRX dividend calendar documentation and our alignment methodology are described in detail on the Methodology page, including the known limitations for fiscal years where KRX trading calendar shifts the ex-date boundary in non-obvious ways.