Director Buying’s First Month: It Lagged, and One Month Is Too Short to Tell

On September 10 we took every stock MoatMap covers, 26,848 of them across 35 markets, and sorted each one by what its directors had done with their own money in the months before: several directors buying, net buying, filings that netted out, net selling, no filings at all, and filings we could not value. We wrote the groups down and froze them. Today the month closed, and the result is on the Decile Tracker.

The summary belongs up front, and it is not the one we hoped for: in its first month, director buying did not help. Stocks where three or more directors had been buying did worse than stocks where nobody filed anything. It is equally important to say what this does not mean. A single 30-day window is far too short to conclude anything about director buying, in either direction, and we explain why below.

The Numbers

These are stocks in the investable scope (above a $200M market cap with at least $1M of daily traded value when the groups were formed), measured in US dollars including dividends, from September 10 to October 10. We lead with the median because a handful of large winners can drag an average a long way.

Investable scope, USD total return, September 10 to October 10, 2026. 96% to 99.7% of each group has a measured return.

Stocks with several directors buying fell 4.3% at the median against 3.1% for stocks with no director filings, and fewer of them rose: 32% against 38%. Net director selling was the best group of the six, at -2.4%. Across every stock we cover, with no size floor, the order at the median is the same: cluster buying -4.3%, no filings -3.3%.

The month itself was unusual. The S&P 500 rose 3.0% and the MSCI All Country index 1.3%, yet the median stock we cover fell about 3% in dollars. The largest companies carried the indexes while most others slipped, and director buying clusters in smaller companies, which is where the falls were.

Why One Month Is Too Short to Draw Conclusions

We want to be plain about this, because a table like the one above is easy to over-read in either direction.

• The horizon is wrong for the signal. The research on insider purchases measures their effect over six to twelve months, not four weeks. A director who buys is usually expressing a view about the next year or two, not the next month.
• One market regime is one data point. A month in which large companies rose and most others fell says more about that month than about any signal that leans toward smaller companies.
• The groups are small where it matters. 327 investable stocks had three or more directors buying. A gap of about one percentage point between medians in a group that size can come and go with the next month’s news.
• The record is young. This is the first scored month for the insider groups and the third for the decile tracker as a whole. The tracker grades its own evidence, and today that grade is “early evidence, not proof”.

So the honest reading is not “director buying does not work”. It is “director buying did not work this month, and this month cannot tell us whether it works”. A good month next time would not prove the opposite either. What would mean something is a pattern that holds across many months and different markets, and that takes time to build.

The Same Month for StockRank

The StockRank deciles were measured over the same window, on the same stocks. Decile 1 holds the stocks the model liked most on September 10, decile 10 the ones it liked least.

Investable scope, USD total return, September 10 to October 10, 2026.

In a month when the typical stock fell, the best-scored tenth fell least and the worst-scored tenth fell most, close to step by step down the ladder, and more than half of the top tenth rose. Quality, Value and Momentum each pointed the same way on their own. That is encouraging, and it is exactly as much one month as the insider result is. Our first decile month looked quite different, with momentum fully reversed. Neither month settles anything alone.

Why We Publish It Anyway

We said before the month began that we would publish whatever this window showed. A record that only reports its good months is a backtest with extra steps. The groups were formed on live data, on a published date, before the returns existed, and once scored they are frozen: the only way a published number can change is a logged restatement.

The October groups are already formed, 27,579 stocks this time, and they get their result on November 10. That post will be written the same way, whichever way it goes. Every month of the record is on the Decile Tracker, the method is on the methodology page, and the filings behind the insider groups are in the all-markets insider feed. For how buybacks and director buying combine, see the double signal.

Research, not investment advice. Past returns, and especially a single month of them, say little about future returns.