How to Get SEC Form 4 Insider Trading Data in Python
August 10, 2026
Insider trading data comes from SEC Forms 3, 4 and 5, the ownership reports that officers, directors and large shareholders file under Section 16 of the Securities Exchange Act of 1934. Form 4 is the one that tracks trades, and it must be filed “before the end of the second business day following the day on which a transaction resulting in a change in beneficial ownership has been executed” (sec.gov Form 4, read 10 August 2026). Every filing is public on EDGAR as an XML document. In Python, xfl.insiders("NVDA", period="1y") returns one row per transaction with the SEC transaction code already decoded, and that decoding is what decides whether the analysis holds up.
What are Forms 3, 4 and 5?
Section 16 obliges a director, an officer, or a beneficial owner of more than 10% of a registered class of equity securities to report both holdings and trades. Three forms carry the obligation, and each has its own clock.
Form 3 is the initial statement of what an insider already owns, filed “within 10 days after the event by which the person becomes a reporting person” (sec.gov Form 3, read 10 August 2026). Form 4 reports each subsequent change in beneficial ownership on the two-business-day clock quoted above. Form 5 sweeps up transactions that were exempt from Form 4 or were simply missed, and is filed “on or before the 45th day after the end of the issuer’s fiscal year” (sec.gov Form 5, read 10 August 2026).
That two-day deadline is the reason Form 4 is worth reading at all. Compare it with a Form 13F, the quarterly institutional holdings report explained in the 13F guide: a 13F arrives up to 45 days after the quarter it describes and names funds rather than people. A Form 4 names the person, the trade date, the share count and the price, usually within the week.
Why is the raw EDGAR feed hard to use for analysis?
Access is not the obstacle. Shape is.
Each Form 4 is a separate XML document, and one document can carry several transaction lines split across two tables, one for ordinary shares and one for derivatives. Building a panel means walking the daily filing index, fetching every document, and flattening those tables into rows. The SEC caps automated access at “10 requests/second” and asks callers to declare a User-Agent header carrying a company name and contact address (sec.gov, Accessing EDGAR Data, read 10 August 2026), so the exercise is a rate-limited crawl rather than a download.
A bulk route exists. The SEC publishes Insider Transactions Data Sets extracted from Forms 3, 4 and 5, covering January 2006 to June 2026, and states that “the data sets will be updated quarterly” (sec.gov, read 10 August 2026). That removes the crawl and adds two constraints in its place: history begins in 2006, and the current quarter is absent until the next posting. The same tradeoff appears in fundamentals, and the EDGAR API comparison works through it in more detail.
Either route leaves the harder problem untouched. The filing carries a raw transaction code, a single letter, and nothing that tells a program what the letter means.
How do you pull Form 4 data in Python?
import xfinlink as xfl
xfl.set_api_key("YOUR_API_KEY") # free at https://xfinlink.com/signup
df = xfl.insiders("NVDA", period="1y")
df["date"] = df["transaction_date"].dt.date
print(df[["date", "insider_name", "insider_role", "transaction_code",
"transaction_type", "shares", "transaction_price"]].head(5).to_string(index=False))
print(df["transaction_type"].value_counts().to_string()) date insider_name insider_role transaction_code transaction_type shares transaction_price
2026-08-05 COXE TENCH Director G gift 500000 0.0
2026-06-25 HUDSON DAWN E Director A grant_or_award 1211 0.0
2026-06-25 Dabiri John Director A grant_or_award 1211 0.0
2026-06-25 Neal Stephen C Director A grant_or_award 1211 0.0
2026-06-25 LORA MELISSA Director A grant_or_award 1211 0.0
transaction_type
open_market_sell 361
grant_or_award 24
tax_withholding 23
gift 15
other 5Both the raw transaction_code and a decoded transaction_type sit on every row, alongside the insider name, the role, the acquired-or-disposed flag, the share count, the price, and the holdings figure after the transaction. History runs back to 1996, and form_type separates Forms 3, 4 and 5 when a study needs only one of them. The full field list is in the docs.
Which Form 4 transaction codes matter?
Eight codes account for almost everything a large company files. The SEC descriptions below are quoted from the Form 4 general instructions (sec.gov, read 10 August 2026).
| Code | SEC description | What it means for analysis | Decoded value |
|---|---|---|---|
| P | “Open market or private purchase of non-derivative or derivative security” | The insider paid cash. The signal everyone is looking for. | open_market_buy |
| S | “Open market or private sale of non-derivative or derivative security” | A genuine sale, though often a scheduled one. | open_market_sell |
| A | “Grant, award or other acquisition pursuant to Rule 16b-3(d)” | Compensation. No decision to buy is expressed. | grant_or_award |
| F | “Payment of exercise price or tax liability by delivering or withholding securities incident to the receipt, exercise or vesting of a security issued in accordance with Rule 16b-3” | Shares withheld to settle a tax bill. Mechanical. | tax_withholding |
| M | “Exercise or conversion of derivative security exempted pursuant to Rule 16b-3” | An option turning into stock, not a purchase. | option_exercise |
| G | “Bona fide gift” | A transfer at no price. | gift |
| D | “Disposition to the issuer of issuer equity securities pursuant to Rule 16b-3(e)” | Sold back to the company, not into the market. | sale_to_issuer |
| C | “Conversion of derivative security” | A change of instrument. | conversion_of_derivative |
Codes A, F, M and G describe things that happen to an insider rather than things an insider chooses. They dominate the row count.
What is the most common mistake in insider analysis?
Counting every disposition as a sale. The acquisition_or_disposition flag reads “D” for an open-market sale, for shares withheld to pay tax, for a gift, and for the disposal leg of an option exercise, so a filter on that flag alone sweeps all four into one number.
Across eight large caps over three years, the difference is not marginal.
tickers = ["AAPL", "MSFT", "JPM", "XOM", "KO", "WFC", "GM", "PFE"]
d = xfl.insiders(tickers, period="3y")
sold = d[d["acquisition_or_disposition"] == "D"]
print(sold.groupby("transaction_type")["transaction_value"].agg(["size", "sum"])
.sort_values("sum", ascending=False).to_string())
open_market = sold[sold["transaction_type"] == "open_market_sell"]["transaction_value"].sum()
print(f"every disposition counted as selling: ${sold['transaction_value'].sum()/1e9:,.2f}bn")
print(f"open-market sales only: ${open_market/1e9:,.2f}bn") size sum
transaction_type
open_market_sell 388 2.006649e+09
tax_withholding 508 1.419691e+09
other 2 3.120996e+05
gift 72 0.000000e+00
option_exercise 30 0.000000e+00
every disposition counted as selling: $3.43bn
open-market sales only: $2.01bnThe naive figure is 1.71 times the real one. Tax withholding alone supplies $1.42 billion of the gap, spread over 508 rows, and those shares never reached the market: the issuer withheld them at vest to settle the recipient’s tax liability. A dashboard built on the disposition flag reports a wave of executive selling every time restricted stock vests on schedule.
How do you filter to the transactions that carry signal?
Filter on the decoded type rather than the direction, and add a size floor so that token purchases do not crowd the result.
buys = xfl.insiders(tickers, period="3y",
transaction_type="open_market_buy",
min_value=250_000)
buys["date"] = buys["transaction_date"].dt.date
print(buys[["ticker", "date", "insider_name", "insider_role",
"shares", "transaction_price", "transaction_value"]]
.sort_values("date", ascending=False).to_string(index=False))ticker date insider_name insider_role shares transaction_price transaction_value
PFE 2026-08-05 Buckley Mortimer J Director 37632 25.5200 9.603686e+05
PFE 2026-08-05 BLAYLOCK RONALD E Director 39231 25.4600 9.988213e+05
KO 2025-10-24 LEVCHIN MAX R Director 7206 69.8706 5.034875e+05
KO 2025-10-23 LEVCHIN MAX R Director 4197 70.3062 2.950751e+05
MSFT 2025-04-23 SMITH BRADFORD LEE President 3842 377.4650 1.450221e+06
PFE 2025-02-13 BLAYLOCK RONALD E Director 19457 25.6500 4.990720e+05
GM 2024-07-26 JACOBSON PAUL A CFO 25000 44.1100 1.102750e+06
XOM 2024-06-17 DREYFUS MARIA S Director 18310 109.2510 2.000386e+06
XOM 2023-11-06 UBBEN JEFFREY W Director 50000 105.9882 5.299410e+06
XOM 2023-11-06 UBBEN JEFFREY W Director 50000 105.9872 5.299360e+06
XOM 2023-11-06 UBBEN JEFFREY W Director 150000 105.9510 1.589265e+07Eleven rows out of 1,916. That ratio is the point of the whole exercise. Insiders at large companies are paid in stock, so acquisitions arrive on a vesting calendar and disposals arrive on a tax calendar, while a purchase requires the person to write a cheque against a position they already hold. Scarcity is what gives code P its information content, and averaging it together with the other codes destroys exactly that.
The remaining filters narrow it further. insider_role matches a substring, so insider_role="CEO" or insider_role="Director" separates the officers who see the operating numbers daily from board members who see them quarterly. min_value screens on dollars rather than shares, which keeps the comparison honest across a $25 stock and a $380 one. Setting include_amendments=True brings in corrected filings when the audit trail matters.
FAQ
Is Form 4 data free?
The filings are free on EDGAR, and the SEC’s quarterly data sets are free to download. The xfinlink insiders endpoint is on the paid plans; see pricing for the tiers and their limits.
How quickly does a trade appear after it happens?
Form 4 is due before the end of the second business day after the trade, so most transactions become public within a week of execution. Each row carries both transaction_date and filing_date, which is how the reporting lag can be measured directly rather than assumed.
Do open-market purchases predict returns?
That is an empirical question and the answer varies by horizon, company size and role. The requirement is a clean sample first: a study that treats grants and tax withholding as trades is measuring the compensation calendar, not insider conviction.
Built with xfinlink — free financial data API for Python. pip install -U xfinlink
pip install -U xfinlink