Research
Follow the boss, not the family office.
When a German company's board members buy their own stock, that's a real signal. When their spouses, holding companies, or financial advisors buy — that's noise, and usually negative noise. The gap between the two is enormous and almost nobody is talking about it.
Board members (Vorstand)
54%
positive 6–12 months later · mean return +6.6% · n=183
Related parties (spouses, holding cos)
13%
positive 6–12 months later · mean return -16.1% · n=67
That's ≈4.0× the hit rate. Same country, same time window, same market. The only thing that's different is who pulled the trigger on the trade.
What this means in plain English
- Every EU public company files a notice when a director buys its stock.They also file the same notice when the director's spouse, holding company, or close associate buys. Most data products mash all of these together into one “insider buy” flag.
- That's a mistake.Board members buying their own stock is a real signal — they're the people who actually know how the next quarter is going. Connected-party buys are something else entirely: tax structuring, family office reshuffles, holding-company transfers, occasionally bag-holding for a director who can't buy directly.
- If you're a retail investortrying to copy insider trades from public filings, you're probably losing money on the related-party leg without realising it. The right filter is simple: only follow trades where the filer is an actual board member or executive. Skip everything else.
- If you're a fund, the role tag is the moat. The 5-minute alert isn't doing the work; the classification is.
BaFin DE sample, 680 board-equivalent buys filed between 2025-04-17 and 2026-04-24. Returns measured to the most recent stored close (last refreshed 2026-05-21). Winsorised at 1st/99th percentile.
Returns by role × holding period (DE buys)
The same window applies to every row inside a holding-period bucket, so the spread between roles is alpha by construction — both buckets ate the same market over the same days. No external benchmark required to interpret the gap.
| Holding | Vorstand | Aufsichtsrat | Related party | Other exec |
|---|---|---|---|---|
| <30d | +8.0% n=26 · hit 88% | +10.3% n=11 · hit 82% | +5.8% n=19 · hit 79% | +18.3% n=4 · hit 100% |
| 30-90d | -1.3% n=64 · hit 45% | +3.9% n=12 · hit 67% | -6.0% n=61 · hit 25% | -0.6% n=6 · hit 50% |
| 90-180d | +9.8% n=66 · hit 64% | +11.6% n=34 · hit 50% | -13.3% n=39 · hit 23% | -19.2% n=2 · hit 0% |
| 180-365d | +6.6% n=183 · hit 54% | -6.0% n=71 · hit 49% | -16.1% n=67 · hit 13% | -0.5% n=11 · hit 36% |
| 365d+ | +24.5% n=4 · hit 100% | — | — | — |
Mean return shown above each cell; n and hit-rate (% positive) below. “Vorstand” is the German PDMR Management Board role. “Aufsichtsrat” is the Supervisory Board. “Related party” (in enger Beziehung) is the German MAR Article 19 designation for spouses, holding companies, and other legal persons closely associated with a PDMR.
Vorstand minus Related-party, within bucket
Reading off the matrix: at every horizon, board-level buys outperform connected-party buys. The hit-rate spread is even larger than the mean-return spread, which matters because hit-rate is robust to outliers. At 6 to 12 months held, board buys are positive 54.9% of the time; related-party buys, 13.4%.
| Holding | Sample | Mean spread | Hit-rate spread |
|---|---|---|---|
| <30d | Vorstand n=26 · Related n=19 | +2.2pp(+8.0% − +5.8%) | +10pp(88% vs 79%) |
| 30-90d | Vorstand n=64 · Related n=61 | +4.7pp(-1.3% − -6.0%) | +21pp(45% vs 25%) |
| 90-180d | Vorstand n=66 · Related n=39 | +23.1pp(+9.8% − -13.3%) | +41pp(64% vs 23%) |
| 180-365d | Vorstand n=183 · Related n=67 | +22.7pp(+6.6% − -16.1%) | +41pp(54% vs 13%) |
A note on what this is and is not: it does notclaim a tradable absolute return — that needs a benchmark we're not showing. It claims that conditional on holding period, the role of the filer is highly predictive of the eventual outcome.
The signal grows with holding period
A common assumption for insider data is that the signal decays quickly — “real-time alerts or nothing.” In our data the opposite is true: the gap between board buys and related-party buys widensas you hold longer. At 30 days the spread is a couple of points; at 90–180 days it's 25 percentage points; the hit-rate gap goes from +10pp to +42pp. Whatever the board insiders are seeing plays out over months, not minutes.
This matches the published academic finding that insider-buy alpha accumulates over a 6–18 month window (Lakonishok & Lee, Cohen-Malloy-Pomorski). Our European sample replicates the shape, with a sharper role-quality split than the US literature, likely because EU disclosure separates “Vorstand” and “in enger Beziehung” cleanly where US Form 4 doesn't.
Methodology
- Sample: 680 BaFin DE PDMR buys with both a disclosed transaction price and a current-price snapshot. Roles classified by the German MAR Article 19 designation (Vorstand, Aufsichtsrat, in enger Beziehung, Sonstige Führungsperson). Filings dated 2025-04-17 to 2026-04-24.
- Returns: measured from the disclosed transaction price to the most recent stored close (last refresh 2026-05-21). Holding period is calendar days from transaction to refresh, not trading days.
- Outliers: winsorised globally at the 1st/99th percentile (clipped to [-58.3%, +95.5%]) so a small number of extreme outcomes don't drive the bucket means.
- No external benchmark. Within-bucket comparison holds the market constant by construction: every observation in a holding-period bucket spent the same number of days exposed to the same market. That makes role spreads benchmark-independent. We do notpublish absolute alpha until we've wired in a survivorship-corrected price feed (in-progress, EODHD-class).
- Single-jurisdiction.The role taxonomy is BaFin-specific; pan-EU normalisation is in flight but currently this page is a DE study. Don't generalise to SE / NO / IT until we publish those separately.
- What we deliberately did not include: cluster-buy analysis (n ≈ 18 in our current sample, too thin to test against the US literature on thousands of cluster events); insider-sell analysis (cells too sparse per role/horizon to interpret cleanly); absolute return vs. STOXX 600 (deferred until prices are sourced from a survivorship-aware feed).
Coverage matrix →
Per-country tier and freshness for the underlying feed.
API access →
PDMR + 5%-threshold whales over 23+ regulators. Trial key in 30 seconds.
Live leaderboard →
Per-insider 3m / 12m forward returns, refreshed nightly.