Does following company insiders pay?
When a company's executives, directors or largest shareholders trade its shares, they must report it to the SEC, usually within two business days. We took every one of those reports for 3,815 US companies since 2010, 1,704,474 in all, and asked a simple question: if you had bought the shares the day after each report became public, would you have beaten the S&P 500?
The short answer
Not reliably. Shares that insiders bought were ahead of the S&P 500 by +1.5% on average a year after the filing became public; shares they sold without a plan were behind it by 0.7%. The S&P 500 is weighted towards the largest companies and most companies here are smaller, so we also compare with routine share awards, which involve no decision by the insider (−1.0% against the index over 12 months). Against that yardstick, buying after insiders bought was +0.6% over 12 months (t = 0.57), and none of the gaps against awards is large enough to rule out chance. Insiders themselves did better than their followers: part of the move came before the filing was public.
- When insiders bought shares with their own money after 3 months +0.3%, after 12 months +1.5% vs the S&P 500 No reliable evidence
- When they sold without a pre-set plan after 3 months −0.2%, after 12 months −0.7% vs the S&P 500 No reliable evidence
- When they sold under a pre-set 10b5-1 plan after 3 months +0.4%, after 12 months +1.2% vs the S&P 500 No reliable evidence
- When they were simply granted shares or options after 3 months −0.7%, after 12 months −1.0% vs the S&P 500 No reliable evidence
- When they exercised options after 3 months −0.4%, after 12 months −0.7% vs the S&P 500 No reliable evidence
- When shares were withheld to pay tax after 3 months −0.9%, after 12 months −0.5% vs the S&P 500 No reliable evidence
Buys vs sells vs awards vs exercises
Average return over and above the S&P 500, measured from the close on the first trading day after each filing. "vs awards" compares each month's events with award filings made public in the same month, over the same 12 months, a fairer baseline for the kind of companies involved; it is the average of those monthly gaps, with its t-statistic.
| What the insider did | Events | vs S&P 500, 3 months | vs S&P 500, 12 months | Typical (median), 12 months | vs awards, 12 months | Beat the S&P 500 | Evidence vs S&P 500 |
|---|---|---|---|---|---|---|---|
| Open-market buys (P) | 68,406 | +0.3% | +1.5% | −5.1% | +0.6% t = 0.57 | 43.3% | No reliable evidence t = 0.17 |
| Open-market sales, all (S) | 408,824 | +0.0% | +0.1% | −4.5% | +1.1% t = 0.66 | 44.0% | No reliable evidence t = 0.49 |
| Sales, discretionary | 223,193 | −0.2% | −0.7% | −4.2% | −0.2% t = −0.18 | 43.9% | No reliable evidence t = −0.25 |
| Sales under a 10b5-1 plan | 185,631 | +0.4% | +1.2% | −5.1% | +3.0% t = 1.20 | 44.1% | No reliable evidence t = 1.07 |
| Stock and option awards (A) | 535,815 | −0.7% | −1.0% | −4.5% | baseline | 43.4% | No reliable evidence t = −0.11 |
| Option exercises (M) | 325,622 | −0.4% | −0.7% | −4.5% | 0.0% t = 0.00 | 43.5% | No reliable evidence t = −0.12 |
| Shares withheld for tax (F) | 365,807 | −0.9% | −0.5% | −4.6% | +0.7% t = 1.20 | 43.5% | No reliable evidence t = 0.33 |
What insiders earned vs what you would get
Insiders trade before anyone knows. Measured from the day of the trade itself, their purchases were +1.8% against the S&P 500 after 3 months; a follower buying after the filing got +0.3%. The gap is the move that happens before the news is public.
| What the insider did | Insider, 3 months | Follower, 3 months | Insider, 12 months | Follower, 12 months |
|---|---|---|---|---|
| Open-market buys (P) | +1.8% | +0.3% | +2.8% | +1.5% |
| Open-market sales, all (S) | −0.1% | +0.0% | +0.0% | +0.1% |
| Sales, discretionary | −0.5% | −0.2% | −0.9% | −0.7% |
| Sales under a 10b5-1 plan | +0.3% | +0.4% | +1.2% | +1.2% |
| Stock and option awards (A) | −0.6% | −0.7% | −0.9% | −1.0% |
| Option exercises (M) | −0.4% | −0.4% | −0.7% | −0.7% |
| Shares withheld for tax (F) | −0.9% | −0.9% | −0.6% | −0.5% |
Does it matter who is trading?
Open-market buys and sales by role, average return against the S&P 500 after the filing. Chief executives and chief financial officers are identified from the job title on the filing.
| Role | Buys | Sales | |||||
|---|---|---|---|---|---|---|---|
| Events | 3 months | 12 months | Evidence | Events | 12 months | Evidence | |
| Chief executive | 11,008 | −0.3% | +0.6% | No reliable evidence | 61,218 | +0.6% | No reliable evidence |
| Chief financial officer | 3,655 | +1.2% | +4.7% | No reliable evidence | 34,355 | +0.2% | No reliable evidence |
| Other officers | 10,749 | +0.4% | +0.6% | No reliable evidence | 198,837 | +0.2% | No reliable evidence |
| Directors | 34,060 | +0.6% | +3.3% | No reliable evidence | 94,330 | −0.4% | No reliable evidence |
| 10% owners | 7,199 | −0.2% | −3.4% | No reliable evidence | 15,914 | −0.7% | No reliable evidence |
Cluster buys: when several insiders buy at once
A cluster is three or more different insiders buying the same company's shares on the open market within 30 days, a signal often said to be stronger than a lone purchase. We found 4,300 clusters. After 3 months they were +0.7% against the S&P 500 and after 12 months +2.2%, against +1.4% for other insider buys. Evidence that clusters beat the S&P 500: no reliable evidence.
| Signal | Events | 1 month | 3 months | 6 months | 12 months | Beat the S&P 500 (12 months) |
|---|---|---|---|---|---|---|
| Cluster buys | 4,300 | +0.1% | +0.7% | −0.3% | +2.2% | 43.9% |
| Other insider buys | 64,106 | +0.1% | +0.3% | −0.4% | +1.4% | 43.3% |
A portfolio that follows insiders
At the end of every month, buy an equal amount of each company where insiders were net buyers on the open market in the previous three months (about 290 companies at a time), and hold for a month. From April 30, 2010 this returned +13.3% a year, against +13.9% for the S&P 500 and +10.8% for an equal stake in every company we cover. It beat the typical company, but not the index, with bigger swings than the S&P 500. After allowing for its market exposure (beta 1.11), its return relative to the S&P 500 (alpha) was −1.1% a year.
| Strategy | 2010–2026 | 2010–2017 | 2018–2026 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Return a year | Swing a year | Sharpe | Worst fall | Beta | Return a year | Sharpe | Return a year | Sharpe | |
| Insiders buying (long) | +13.3% | 19.1% | 0.67 | −35.9% | 1.11 | +17.8% | 1.17 | +9.5% | 0.40 |
| Insiders selling, discretionary (long) | +12.3% | 17.3% | 0.67 | −28.1% | 1.10 | +14.5% | 1.02 | +10.3% | 0.46 |
| Buyers minus sellers | +1.1% | 5.9% | 0.21 | −16.5% | 0.01 | +2.8% | 0.63 | −0.4% | −0.02 |
| All covered companies, equal weight | +10.8% | 17.4% | 0.59 | −29.8% | 1.08 | +13.8% | 0.97 | +8.2% | 0.36 |
| S&P 500 (SPY) | +13.9% | 14.3% | 0.87 | −23.9% | 1.00 | +13.4% | 1.11 | +14.2% | 0.74 |
Did it hold up over time?
The same event test split into two halves. A real edge should show up in both; one that flips sign is more likely luck or a passing market phase.
| What the insider did | 2010–2017 | 2018–2026 | ||
|---|---|---|---|---|
| 3 months | 12 months | 3 months | 12 months | |
| Open-market buys (P) | +1.1% | +2.4% | −0.1% | +0.9% |
| Sales, discretionary | +0.2% | +1.9% | −0.7% | −3.3% |
| Sales under a 10b5-1 plan | +1.3% | +6.2% | −0.1% | −1.8% |
| Stock and option awards (A) | +0.5% | +2.3% | −1.4% | −3.1% |
Does it matter how many are buying?
Each month from April 30, 2010 to October 30, 2026 (198 months), we counted how many insiders were net buyers or net sellers of each company over the previous 91 days, using only reports dated before the month end, and held those companies for a month. In a month with fewer than 3 such companies a row holds cash at the Treasury bill rate, so every row covers the same months. Companies with 2 or more insiders buying returned +12.6% a year, against +12.0% for those with 3 or more selling, +10.8% for every covered company on the same rules and +13.9% for the S&P 500. The typical company held with 2 or more insiders buying was worth about $747m, and with 3 or more selling about $5.1bn, against $2.2bn for the typical covered company. Against all covered companies in the months it held shares, 2 or more buying was +2.1% a year (t = 1.53); against companies of similar market value it was +1.7% (t = 1.16).
| Companies with… | Months invested | Held (average) | Return a year | Against all companies | Evidence | Against similar-size companies | Sharpe | Beta | Biggest drop |
|---|---|---|---|---|---|---|---|---|---|
| 1 or more insiders net buying | 198 of 198 | 292 | +13.5% | +2.7% | Strong evidence t = 2.63 | +2.2% t = 2.17 | 0.68 | 1.10 | -34.9% |
| 2 or more net buying | 198 of 198 | 138 | +12.6% | +2.1% | Some evidence t = 1.53 | +1.7% t = 1.16 | 0.64 | 1.10 | -36.1% |
| 3 or more net buying | 198 of 198 | 69 | +11.7% | +1.4% | No reliable evidence t = 0.82 | +1.6% t = 0.84 | 0.58 | 1.11 | -39.4% |
| 5 or more net buying | 198 of 198 | 20 | +11.2% | +1.0% | No reliable evidence t = 0.39 | +3.5% t = 1.03 | 0.54 | 1.04 | -30.7% |
| 1 or more insiders net selling | 198 of 198 | 872 | +12.2% | +1.3% | Some evidence: sellers wrong t = 1.92 | +0.2% t = 0.42 | 0.67 | 1.10 | -27.8% |
| 2 or more net selling | 198 of 198 | 548 | +12.4% | +1.4% | Some evidence: sellers wrong t = 1.67 | +0.1% t = 0.19 | 0.68 | 1.10 | -29.3% |
| 3 or more net selling | 198 of 198 | 349 | +12.0% | +1.1% | No reliable evidence t = 1.02 | −0.3% t = −0.29 | 0.65 | 1.11 | -31.8% |
| 5 or more net selling | 198 of 198 | 146 | +11.2% | +0.5% | No reliable evidence t = 0.33 | −0.9% t = −0.69 | 0.61 | 1.10 | -34.2% |
| All companies, equal weight | 198 of 198 | +10.8% | 0.59 | 1.08 | -29.8% | ||||
| S&P 500, buy and hold | 198 of 198 | +13.9% | 0.87 | 1.00 | -23.9% |
Can a smarter insider strategy beat the market?
Following methods published by Quiver Quantitative, we rebuilt two insider strategies on our own data from 2010: a monthly ranking of companies by insider purchases (five ways of ranking them, on the 500 largest companies we cover and on every one worth at least $300m), and a model that scores each purchase on the insider’s record, the size of the trade, how many colleagues are buying, momentum and value, then holds the ten highest-scoring companies week by week. We added 10 variants of our own, fixed before looking at the results, so 21 strategies in all. Every figure is after 10 basis points per trade unless stated.
None reached a Sharpe ratio above 1. The best full-period Sharpe ratio was 0.68 (0.74 with no costs, 0.59 at 25 basis points), against 0.75 for simply holding the S&P 500; 0 of 21 beat that. Most of what these strategies earned was the market itself: with the market’s ups and downs hedged out, the best Sharpe ratio was 0.13. Picking the strategy with the best Sharpe ratio on 2010–2017 data alone, dollar value, above 200-day average, 500 largest, 25 holdings, traded next day, gives 0.98 in those years (the S&P 500: 0.92) but 0.47 from 2018 (the S&P 500: 0.65): the edge did not survive. The conviction model’s scores did not separate winners from losers out of sample: two months after the report, purchases in its top tenth were −0.1% against the S&P 500 on average and those in its bottom tenth −0.6%.
| Strategy | Return a year | Sharpe | Biggest drop | Beta | Hedged Sharpe | Sharpe 2010–17 | Sharpe 2018–26 | Sharpe Aug 2021–Aug 2026 | Turnover a year |
|---|---|---|---|---|---|---|---|---|---|
| Quiver method A (replicated) | |||||||||
| Dollar value bought, 500 largestfrom April 30, 2010 | +10.4% | 0.49 | −50% | 1.07 | −0.31 | 0.68 | 0.36 | 0.370.55 with no cash rate | 444% |
| Cluster (distinct insiders), 500 largestfrom April 30, 2010 | +11.4% | 0.53 | −48% | 1.07 | −0.24 | 0.71 | 0.39 | 0.210.39 with no cash rate | 514% |
| Officers only, dollar value, 500 largestfrom April 30, 2010 | +9.5% | 0.44 | −45% | 1.11 | −0.37 | 0.61 | 0.33 | 0.040.20 with no cash rate | 467% |
| Stake increase, 500 largestfrom April 30, 2010 | +7.8% | 0.37 | −53% | 1.11 | −0.48 | 0.49 | 0.30 | 0.290.46 with no cash rate | 581% |
| Dollar value, above 200-day average, 500 largestfrom April 30, 2010 | +11.6% | 0.57 | −48% | 0.99 | −0.22 | 0.85 | 0.39 | 0.370.56 with no cash rate | 595% |
| Dollar value bought, all $300m+from April 30, 2010 | +8.3% | 0.36 | −59% | 1.16 | −0.26 | 0.57 | 0.26 | 0.190.30 with no cash rate | 451% |
| Cluster (distinct insiders), all $300m+from April 30, 2010 | +8.8% | 0.40 | −43% | 1.08 | −0.30 | 0.60 | 0.26 | 0.160.30 with no cash rate | 552% |
| Officers only, dollar value, all $300m+from April 30, 2010 | +10.2% | 0.43 | −55% | 1.16 | −0.22 | 0.56 | 0.37 | 0.250.37 with no cash rate | 461% |
| Stake increase, all $300m+from April 30, 2010 | +8.3% | 0.37 | −59% | 1.16 | −0.27 | 0.51 | 0.29 | 0.370.49 with no cash rate | 581% |
| Dollar value, above 200-day average, all $300m+from April 30, 2010 | +13.3% | 0.56 | −45% | 1.08 | −0.07 | 0.63 | 0.53 | 0.470.61 with no cash rate | 616% |
| Quiver method B (replicated) | |||||||||
| Conviction model, all $300m+from January 7, 2013 | +6.2% | 0.29 | −65% | 1.21 | −0.37 | 0.38 | 0.25 | 0.050.18 with no cash rate | 1242% |
| Our variants | |||||||||
| Dollar value bought, 500 largest, traded next dayfrom April 1, 2010 | +12.2% | 0.57 | −47% | 1.07 | −0.19 | 0.89 | 0.34 | 0.470.65 with no cash rate | 442% |
| Dollar value bought, 500 largest, 25 holdings, traded next dayfrom April 1, 2010 | +11.5% | 0.56 | −45% | 1.07 | −0.33 | 0.86 | 0.36 | 0.260.45 with no cash rate | 425% |
| Dollar value, above 200-day average, 500 largest, traded next dayfrom April 1, 2010 | +11.7% | 0.59 | −41% | 0.97 | −0.21 | 0.93 | 0.36 | 0.310.50 with no cash rate | 588% |
| Dollar value, above 200-day average, 500 largest, 25 holdings, traded next day chosen on 2010–2017from April 1, 2010 | +12.7% | 0.68 | −43% | 0.93 | −0.10 | 0.98 | 0.47 | 0.400.63 with no cash rate | 570% |
| Dollar value bought, all $300m+, traded next dayfrom April 1, 2010 | +9.8% | 0.39 | −76% | 1.08 | −0.06 | 0.66 | 0.29 | −0.21−0.08 with no cash rate | 451% |
| Dollar value bought, all $300m+, 25 holdings, traded next dayfrom April 1, 2010 | +12.7% | 0.52 | −50% | 1.12 | −0.05 | 0.74 | 0.42 | 0.140.28 with no cash rate | 424% |
| Dollar value, above 200-day average, all $300m+, traded next dayfrom April 1, 2010 | +13.3% | 0.48 | −60% | 1.00 | 0.08 | 0.72 | 0.43 | 0.050.18 with no cash rate | 610% |
| Dollar value, above 200-day average, all $300m+, 25 holdings, traded next dayfrom April 1, 2010 | +15.7% | 0.66 | −44% | 0.98 | 0.13 | 0.96 | 0.54 | 0.340.51 with no cash rate | 594% |
| Conviction model, 500 largestfrom February 25, 2013 | −2.5% | 0.00 | −66% | 0.99 | −0.56 | 0.19 | −0.07 | −0.13−0.02 with no cash rate | 845% |
| Conviction model, all $300m+, above 200-day averagefrom January 7, 2013 | +2.2% | 0.14 | −67% | 0.98 | −0.50 | 0.76 | −0.09 | −0.38−0.25 with no cash rate | 1756% |
| S&P 500, buy and holdfrom April 30, 2010 | +13.9% | 0.75 | −34% | 1.00 | 0.92 | 0.65 | 0.590.81 with no cash rate | ||
| 500 largest, equal weightrebalanced monthly | +13.1% | 0.69 | −40% | 1.01 | 0.93 | 0.52 | 0.42 | ||
How the test works and its limits
- Data. Every Form 4 in the SEC's free Insider Transactions Data Sets from January 4, 2010 to September 30, 2026, topped up with filings we collect directly to November 2, 2026. Only companies we cover with a clean price history count (4,743).
- No hindsight. Each trade counts from the close on the first trading day after the filing date, when the public could act on it, not from the trade date. The "insider" figures measure from the trade date to show what is lost by waiting.
- Returns include dividends and are compared with SPY, an S&P 500 fund, over 1, 3, 6 and 12 months (21, 63, 126 and 252 trading days). The most recent 12-month results run to filings made in October 30, 2025.
- What counts. One event per insider per filing per type. Purchases (code P) and sales (code S) need a price above zero and a value of at least $10,000. When one trade is reported separately by several related entities (a fund's partnerships, or 10% holders reporting the same shares), with the same company, day, number of shares and value, it counts once; directors and officers always count as separate people, even when they buy identical amounts (3,117 duplicate reports merged). Shares bought within a week of a company's first day of trading are usually allocations in the flotation, not open-market purchases, and are left out (2,939 purchases).
- Prices. Shares that traded under $5 at entry are left out, judged on the price actually paid that day, not today's figure restated for later splits. A closing price that doubles or halves for a single day and is back within 10% the next is treated as a bad record and replaced by the previous close (148 prices). Nothing else is removed, so big winners and big losers count alike.
- Planned sales. A sale is "planned" when the filing ticks the Rule 10b5-1 box (from April 2023) or a footnote of the filing mentions a 10b5-1 plan. Before 2023 the footnote was voluntary, so some planned sales count as discretionary.
- Evidence. Many insiders trade in the same weeks, so their results are not independent. We average all events in each month first and test those monthly averages, allowing for the overlap of 3- to 12-month windows (Newey-West). A t-statistic of about 2 or more is strong evidence; below 1.5 the result could easily be chance.
- Survivorship. The companies are those we cover today. Firms that went bust or were taken over since 2010 are missing, which flatters every row equally but means the absolute numbers are not what an investor at the time could have earned.
- Costs. No trading costs, bid-ask spreads or tax. Many insider buys are in small companies where these are highest, so a real follower would do worse than shown.