← Blog

Data exploration

Stock Price Behavior Around Ex-Dividend Dates

A validation note on the event-window labels, weekly source cadence, and exact-date selection that currently prevent an honest public price-effect conclusion.

Status
Under validation
Sample
528040 intervals audited
Reviewed
Reviewed 29 July 2026

Abstract

This note audits whether the current data can support an event study of stock-price movement around dividend dates. The intended analysis follows prices before and after an event, but the reviewed observations are weekly: 99.6% of consecutive intervals are 6–8 calendar days apart. The current exact-date selection also retains only 18.1% of the ordinary dividend records reviewed.

Validation finding

The Day 0/Day 1 labels do not describe daily observations in this snapshot. The price-effect series and its market conclusions are therefore withheld.

1. Introduction

An event-window study needs a precise anchor: the last eligible close before a stock trades without the upcoming dividend, followed by correctly ordered daily sessions. The existing analysis was designed around that familiar paper structure, with a common baseline and observations on either side of the dividend date.

Before presenting a static version of that analysis, this note tests whether the stored dates actually support the event-day labels and whether the event selection represents the reviewed dividend records.

Validation questions

  • Are consecutive price observations daily or weekly?
  • Do Day 0 and Day 1 identify the sessions described by the analysis?
  • How many ordinary dividend records survive the exact-date match?
  • Can a price-effect chart be interpreted honestly from this snapshot?

2. Methodology audit

The audit examines the spacing between consecutive stored price observations and the result of matching ordinary dividend dates directly to those observations. It does not use or reproduce the private price-effect curves.

Event-day definitions under review

Page definition
Day 0 is described as the last eligible day; Day 1 as the ex-dividend date.
Chart marker
The same interface places its ex-dividend reference at Day 0.
Observed cadence
The median gap between stored observations is 7 days.

These labels conflict with each other, and neither can turn adjacent weekly observations into adjacent trading days. The exact-date audit separately measures how much of the dividend-record set is retained by the current join.

3. Validation Results

The reviewed snapshot supports two data-quality results. Together they are sufficient to stop publication of the intended price-effect chart.

Intervals audited
528040
Median gap
7 days
Exact-date matches
18.1%
Price-effect result
Withheld

Figure 1

Stored observations follow a weekly cadence

528040 intervals

Distribution of calendar-day gaps between consecutive stored stock price observations

6–8 days 99.6%
525778 intervals
9–15 days 0.3%
1666 intervals
16 days or more 0.1%
596 intervals
99.6% of intervals are 6–8 days apart; the median gap is 7 days. Adjacent points should therefore be read as weekly observations, not trading-day offsets.

Figure 2

Exact-date matching retains a narrow subset

18.1% matched

780 of 4302 ordinary dividend records had an exact-date price match.

4302 ordinary dividend records reviewed

780 exact-date matches retained

Requiring a dividend date to equal a stored weekly price date retains 780 of 4302 reviewed records. The excluded records may differ systematically from the retained subset.

Price-effect result withheld

No static price curves, yield-group comparisons, recovery patterns, or event-day effects are published. The current snapshot cannot support those interpretations.

4. Discussion

The cadence result changes the unit of the event window. A point labelled “next trading day” is generally about a week later in the reviewed data. Separately, the exact-date rule keeps fewer than one in five reviewed ordinary dividend records. A curve built from that subset would not answer the daily event question posed by the original paper.

4.1 Limits of the current snapshot

  • Adjacent stored observations are predominantly weekly, so they cannot be interpreted as adjacent trading days.
  • The interface’s Day 0 and Day 1 descriptions conflict with the plotted reference marker.
  • Exact-date matching retains only 18.1% of reviewed ordinary dividend records and may introduce selection bias.
  • The current data do not support a daily price-effect, recovery, prediction, trading, or investment conclusion.

4.2 Release gate

  • Use daily, exchange-calendar-aligned observations and explicitly identify the last eligible close and the ex-dividend session.
  • Replace the exact-date selection with a reviewed event-calendar alignment, then reconcile every included and excluded ordinary dividend record.
  • Review duplicate events and corporate actions before calculating any aggregate price movement.
  • Account for repeated companies, overlapping windows, and broader market movement before interpreting a dividend-date pattern.

Publication rule

The price-effect section remains withheld until the event labels, daily cadence, and selection audit agree. Passing that gate would permit a new review; it would not guarantee a market conclusion.

5. Conclusion

01

The cadence is weekly

99.6% of audited intervals are 6–8 days apart, with a 7-day median.

02

The retained subset is narrow

Exact-date matching keeps 18.1% of the reviewed ordinary dividend records.

03

The market result is not publishable

The reviewed evidence supports a validation decision, not a price, recovery, prediction, trading, or investment conclusion.

This static note documents a withheld result. It is not financial or investment advice.