Research / April 2026
A full-methodology backtest: do insider buys from newly-appointed executives outperform?
This page is a full-methodology backtest of whether Form 4 insider buys from recently-appointed executives
(identified via 8-K Item 5.02 filings within the prior 90 days) beat buys from long-tenured C-suite peers at
stable companies. One year of SEC filings, a fixed cohort definition set before running, and every number below
is reproducible from public EDGAR and Polygon data.
The short answer is no — the headline edge is within the noise floor. The longer answer, with
the two sub-findings that are worth keeping an eye on, is below.
Published 2026-04-17 · Test window 2025-04-16 → 2026-04-16 ·
← All research
+0.82%
90-day mean excess-return difference: new-exec cohort vs. stable-C-suite placebo.
(N=58 vs N=885; p=0.824)
Mean excess return vs. SPY, by cohort
Mean excess return vs. SPY — 30-day horizon
Cohort 1 — new exec (all) N=85
1.92%
Cohort 1 — role-confirmed N=80
0.23%
Cohort 2 — placebo C-suite N=1084
2.40%
Cohort 3 — baseline N=488
-2.02%
Mean excess return vs. SPY — 60-day horizon
Cohort 1 — new exec (all) N=62
8.23%
Cohort 1 — role-confirmed N=57
7.96%
Cohort 2 — placebo C-suite N=944
0.69%
Cohort 3 — baseline N=488
-2.83%
Mean excess return vs. SPY — 90-day horizon
Cohort 1 — new exec (all) N=58
0.67%
Cohort 1 — role-confirmed N=53
0.16%
Cohort 2 — placebo C-suite N=885
-0.16%
Cohort 3 — baseline N=488
-5.56%
Full cohort tables
Cohort 1 — New executive buy (all matches)
| Horizon | N | Mean ret | Median ret |
Mean excess | Median excess | Win % | Excess win % | Std dev |
| 30d |
85 |
2.96% |
1.27% |
1.92% |
1.22% |
57.65% |
56.47% |
22.78% |
| 60d |
62 |
11.65% |
2.98% |
8.23% |
-1.65% |
58.06% |
40.32% |
82.62% |
| 90d |
58 |
5.83% |
6.43% |
0.67% |
-0.46% |
65.52% |
48.28% |
27.53% |
Cohort 1 — New executive buy (role-confirmed only, high-confidence name match)
| Horizon | N | Mean ret | Median ret |
Mean excess | Median excess | Win % | Excess win % | Std dev |
| 30d |
80 |
1.21% |
0.95% |
0.23% |
1.20% |
55.00% |
56.25% |
17.34% |
| 60d |
57 |
11.38% |
2.87% |
7.96% |
-1.65% |
57.89% |
38.60% |
85.86% |
| 90d |
53 |
5.41% |
6.32% |
0.16% |
-0.59% |
64.15% |
47.17% |
28.33% |
Cohort 2 — Placebo: stable C-suite (no 8-K 5.02 in prior 180 days)
| Horizon | N | Mean ret | Median ret |
Mean excess | Median excess | Win % | Excess win % | Std dev |
| 30d |
1084 |
4.07% |
2.38% |
2.40% |
0.56% |
63.19% |
53.04% |
14.44% |
| 60d |
944 |
4.29% |
2.53% |
0.69% |
-1.51% |
60.49% |
44.17% |
20.75% |
| 90d |
885 |
4.58% |
2.31% |
-0.16% |
-3.05% |
57.85% |
41.69% |
25.27% |
Cohort 3 — Baseline: all qualifying Form 4 P-buys
| Horizon | N | Mean ret | Median ret |
Mean excess | Median excess | Win % | Excess win % | Std dev |
| 30d |
488 |
4.80% |
3.65% |
-2.02% |
-3.18% |
70.29% |
34.63% |
11.89% |
| 60d |
488 |
7.80% |
6.16% |
-2.83% |
-4.07% |
75.41% |
35.45% |
20.80% |
| 90d |
488 |
8.17% |
4.70% |
-5.56% |
-8.49% |
67.62% |
30.12% |
23.39% |
Sub-cohort breakdown (Cohort 1)
By role category
| Role | N (90d) | Mean ret 90d | Mean excess 90d | Win % 90d | Excess win % 90d |
|---|
| CEO |
10 ⚠ |
8.55% |
4.61% |
60.00% |
50.00% |
| CFO |
3 ⚠ |
21.41% |
15.07% |
100.00% |
66.67% |
| Director |
33 |
0.81% |
-4.26% |
63.64% |
48.48% |
| Other C-suite |
12 ⚠ |
13.50% |
7.35% |
66.67% |
41.67% |
By tenure bucket (days since appointment)
| Tenure | N (90d) | Mean ret 90d | Mean excess 90d | Win % 90d | Excess win % 90d |
|---|
| 0-30 |
31 |
9.41% |
3.32% |
67.74% |
51.61% |
| 31-90 |
27 ⚠ |
1.73% |
-2.37% |
62.96% |
44.44% |
Statistical tests
30-day excess return — Cohort 1 (all) vs Cohort 2:
t=-0.19, p=0.845, Δmean=-0.48%, N=85/1084
— role-confirmed variant: t=-1.15, p=0.250, Δmean=-2.17%, N=80/1084
60-day excess return — Cohort 1 (all) vs Cohort 2:
t=0.72, p=0.470, Δmean=7.55%, N=62/944
— role-confirmed variant: t=0.64, p=0.520, Δmean=7.28%, N=57/944
90-day excess return — Cohort 1 (all) vs Cohort 2:
t=0.22, p=0.824, Δmean=0.82%, N=58/885
— role-confirmed variant: t=0.08, p=0.937, Δmean=0.31%, N=53/885
Welch's two-sample t-test with normal approximation for p-values. Treat with care when N<30 on either side.
Methodology
- Universe
- NYSE + Nasdaq (XNYS / XNAS / XASE) active US stocks, price ≥ $2 at transaction. 8614 tickers in the universe snapshot.
- Form 4 filter
- Transaction code
P (purchase), Acquired-Disposed code A.
- Cohort 1 definition
- Buyer named in an 8-K Item 5.02 appointment/promotion at the same CIK within 90 days prior to the transaction. Two variants: "all" (headline) and "role-confirmed" (role≠Unknown AND name-match confidence=high).
- Cohort 2 definition
- C-suite buyer (officer; title contains CEO / CFO / COO / President) with NO 8-K Item 5.02 at the same CIK in the prior 180 days.
- Cohort 3
- All qualifying Form 4 P-buys (baseline). Price pulls sampled to 500 to cap Polygon free-tier calls.
- Forward returns
- Entry = closing price on Form 4 transaction date (not filing date). Exit = closing price at +30 / +60 / +90 calendar days, rolled forward to next trading day for weekends/holidays.
- Excess return
- Ticker return minus SPY return over the same window.
- Sample sizes
-
Form 4 total: 7599 |
after clean: 7599 |
$2 floor: 6688 |
exchange: 5458 |
appointments: 10000 |
cohort 1 all: 92 |
cohort 1 role-confirmed: 86 |
cohort 2: 1197.
- Data sources
- SEC EDGAR (efts.sec.gov full-text search + /Archives/ document pulls). Prices via Polygon daily aggregates (adjusted).
- Caveats
- 8-K Item 5.02 appointee extraction is regex-based and imperfect — role=Unknown entries are retained in cohort 1 "all" but excluded from the role-confirmed variant. Some recent buys are excluded from the 90-day bucket because the forward window extends past the data availability date (2026-07-31).
Generated 2026-04-17.