When Insiders Buy the Dip: Returns After a Selloff

Results by prior drawdown state

We bucketed all 45,046 open-market buys by where the stock was trading relative to its trailing 52-week high at the time of the filing. Three buckets: deep selloff (≥20% below the high), mild pullback (10–20% below), and near highs (less than 10% below).

Absolute, split-adjusted returns anchored at the first close on or after the filing date. Code P only; 10b5-1, derivatives, and superseded rows excluded. Sample: Oct 2023–Feb 2026.

The gradient is clear: the deeper the prior drawdown at the time of the buy, the higher the average forward return at every horizon. Deep-selloff buys outperformed near-highs buys by +10.89% vs +6.34% at 3 months and +14.65% vs +8.97% at 6 months. The 1-month edge (+4.92% vs +2.16%) is also material, suggesting the out-performance begins early rather than being a slow-build effect.

One pattern stands out immediately: the mild-pullback bucket sits between the two extremes on average figures, yet its median 3-month return (+5.10%) is the highest of all three buckets — above even the near-highs median (+2.99%) and well above the deep-selloff median (+1.64%). That tells the real story, which the next section unpacks.

Why the median matters — the skew story

The deep-selloff bucket has the highest average return but the lowest median. That combination is the hallmark of a right-skewed distribution: a minority of large winners pulls the mean far above the typical outcome. In practical terms, the majority of insider dip-buys in deeply beaten-down stocks returned something modest or negative at 3 months, while a smaller number recovered dramatically — 20%, 40%, 60% or more — dragging the mean up to +10.89%.

The mild-pullback bucket has the inverse profile: a middling average (+6.75% at 3 months) but the highest median (+5.10%). That means the typical trade in the mild-pullback bucket actually outperformed the typical trade in the deep-selloff bucket, even though the average says the opposite. When stocks are only moderately beaten up, insider conviction tends to be rewarded more consistently — fewer catastrophic losers, fewer lottery-ticket recoveries, and a tighter distribution of outcomes.

What drives the deep-selloff tail? The most plausible interpretation is company-type selection: stocks that are already 20%+ below their 52-week highs tend to be either genuine turnaround situations or companies in structural decline. In the turnaround cases, an insider buying into the weakness is often right and the stock eventually recovers substantially. In the structural-decline cases, the stock keeps falling and the buy returns a large negative. A broad mean across both scenario types produces a high average driven by the turnaround wins, but the median stays low because decline cases are common.

Bull-market caveat — what these numbers do and do not show

The sample window (October 2023 to February 2026) was broadly a rising market. The S&P 500 gained substantially over this period, and that market-wide drift lifts every bucket in absolute terms. You cannot read these results as "insider buys during deep selloffs always make 10%" — the absolute numbers include the market tailwind that applied to everyone.

What the numbers do show reliably is the relative difference between buckets. All three buckets are measured over the same calendar period, so they share the same market drift. The fact that deep-selloff buys outperformed near-highs buys by roughly 4.6 points at 3 months is a statement about the incremental signal from drawdown context — not about absolute returns in all market conditions.

A proper benchmark-adjusted study would subtract the S&P 500 return over the same window for each individual trade (or a sector-index return for a tighter control), giving an excess-return series that is market-neutral by construction. That refinement is on our roadmap. For now, treat the relative-bucket comparison as the valid finding and the absolute numbers as upper-bound estimates that assume the broad market continues to rise.

A second note: the sample does not span a prolonged bear market or a severe recession. The finding that insider dip-buying concentrates into high-average outcomes may look different in a period where beaten-down stocks stay beaten down because the macro environment does not recover. The relative signal is likely to persist across regimes, but the absolute magnitudes should be treated with care.

Methodology

• Population. Every open-market purchase (transaction code P) in the Form4API database with a matured post-trade return. 45,046 transactions in total.
• Sample window. Filings dated October 2023 to February 2026 (the latest window for which 6-month return horizons have fully elapsed at time of query).
• Exclusions. Derivative transactions (IsDerivative = true), 10b5-1 plan executions, superseded (amended) rows, and rows where the ticker or forward return was null. A small fraction (<1%) with data-quality issues in the prior-price series were also excluded.
• Return definition. Absolute, split-adjusted price return anchored at the first market close on or after the filing date (FilingClose), measured at 21, 63, and 126 trading days (approximately 1, 3, and 6 months). Returns are not market-adjusted or sector-adjusted.
• Prior drawdown. Defined as FilingClose / MAX(Close over the prior 365 calendar days) − 1. This is the stock's percentage decline from its trailing 52-week high at the time the Form 4 was filed. Three buckets: ≤−20% (deep selloff), −10% to −20% (mild pullback), and >−10% (near highs).
• Maturity. Only trades whose return horizon has fully elapsed are counted; the sample is backward-looking and very recent filings are under-represented by construction.
• Returns as fractions. The API stores returns as fractions (e.g., 0.1089 = +10.89%). All percentages in this article have been converted for readability.

Reproduce this yourself

Pull the raw open-market buy data and compute your own drawdown buckets. Start with:

# Fetch open-market buys (code P), plan trades excluded, with return data
curl "https://api.form4api.com/v1/transactions?code=P&exclude_10b5=true&has_returns=true" \
  -H "X-Api-Key: $FORM4API_KEY"

# Each record includes:
# {
#   "transactionCode": "P",
#   "is10b5Plan": false,
#   "filedAt": "2024-03-15",
#   "return1m": 0.0492,    // fraction: 0.0492 = +4.92%
#   "return3m": 0.1089,    // fraction: 0.1089 = +10.89%
#   "return6m": 0.1465,
#   "ticker": "AAPL",
#   "insiderName": "...",
#   ...
# }

To reconstruct the prior-drawdown bucket for each transaction, join against a 52-week high series for each ticker and compute (filingClose / trailing52wHigh) − 1. Classify transactions where that ratio is ≤−0.20 as deep-selloff buys, −0.10 to −0.20 as mild pullbacks, and >−0.10 as near-highs buys. Then compute mean and median returns per bucket to reproduce the table above.

The /docs reference covers the full parameter grammar — codes, date range, return horizons, and export to CSV for larger batch analysis.

Frequently asked questions