Stock-price outcomes for public insider transaction disclosures: a holdings-direction comparison

1. Introduction

Public insider-transaction disclosures make it possible to compare recorded transactions with later market-price histories. This note asks a narrow descriptive question: what did the distribution of later price movements look like in the eligible disclosure cohort?

The static view preserves the paper structure of the source application while fixing the scope to all eligible disclosures. It does not recreate the application’s person-position filter or position-level table. Instead, it includes only reviewed aggregates for four broad role categories; raw position titles and person-level output remain outside the public snapshot.

Research questions

• How were later stock-price outcomes distributed across all eligible rows?
• How did those shares differ between the two analyst-mapped holdings directions?
• How do the same distributions look within four broad recorded-position categories?
• What can—and cannot—be concluded from those aggregate differences?

2. Data & Methodology

2.1 Cohort

The reviewed snapshot contains 16557 disclosure rows with an aligned starting observation and a 24th future weekly price observation. The source period in the paper header describes the disclosure publication-date range; it is not the outcome window.

2.2 Holdings-direction mapping

The project maps 27 distinct reportable transaction labels into “holdings increased” or “holdings decreased.” This is an analyst-defined grouping for comparison, not an official regulatory taxonomy. Each direction includes multiple kinds of reportable events.

2.3 Broad role-category mapping

Recorded regulatory-position labels are grouped deterministically into four broad categories: executive leadership, board leadership, board membership, and other governance. Compound roles follow a fixed hierarchy, so each eligible row enters one category rather than being counted in several.

The mapping covers all 16557 eligible rows, with 0 unmapped. Every category-by-direction view shown below has at least 100 rows. The application’s LLM title classifier is not used for this grouping, and neither classifier output nor raw position titles are shipped in the public snapshot.

Mapping basis

Deterministic grouping of recorded regulatory position labels; the LLM title classifier is not used.

Public categories

4

Coverage

16557 mapped · 0 unmapped

Publication threshold

n ≥ 100 in every shown category and direction

2.4 Outcome bins

For each eligible row, the closing-price change is measured from the transaction-date-aligned weekly observation to the 24th future weekly observation, then placed into one of three fixed bands.

“24 weeks” is shorthand for 24 weekly price observations. The reviewed cadence has a median horizon of 168.0 days. The clock starts from the recorded transaction date, not the later publication date.

3. Results

3.1 Sample overview

Eligible rows

16557

Outcome horizon

24 observations

Band threshold

±5%

Mapped labels

27

3.2 Overall outcome distribution

Figure 1 shows a broadly balanced negative-to-positive distribution across all eligible rows, with a smaller neutral band.

3.3 Holdings-direction comparison

Figure 2 freezes the source application’s all-disclosures view. The holdings-increased group is close to balanced between negative and positive outcomes. The holdings-decreased group has a larger negative share and a smaller neutral share.

3.4 Broad role-category distributions

The tables below apply the same outcome bands within each broad recorded-position category. Denominators are shown for the full category and for each holdings direction. Counts describe eligible disclosure rows, not unique people, companies, or events.

4. Discussion

4.1 Reading the distributions

In the overall cohort, 40.7% of outcomes are negative and 39.7% are positive. Within the holdings-increased group, those shares are 38.7% and 39.4%. Within the holdings-decreased group, they are 47.9% and 40.9%.

These aggregate differences describe the rows that passed the inclusion rules. They do not separate transaction type, company, market period, repeated disclosures, or other factors that may affect the comparison.

The broad role-category tables provide another descriptive breakdown of those same eligible rows. They do not isolate the effect of a role, holdings direction, or disclosure from the other differences between rows.

5. Conclusion

Three observations summarize this fixed, all-disclosures view.