---
title: "Famous Investors Herd on the Way Out, Not on the Way In"
url: "https://economistzone.com/insight/famous-investors-herd-on-the-way-out-not-on-the-way-in/"
author: "Razvan Luca"
published: "2026-09-25"
updated: "2026-09-25"
---

# Famous Investors Herd on the Way Out, Not on the Way In

Of the 47 well-known investors we track who filed for the end of June, 26 owned Alphabet, more than any other stock. During that same quarter seven managers added to it and twelve cut it or sold out, which is a strange kind of agreement.

It made me want to test something economists have argued about since the early nineties. Do professional investors really herd, or does it only look that way because they all own the same big companies?

### **A thirty-year-old test**

In 1992 Josef Lakonishok, Andrei Shleifer and Robert Vishny published a simple way to check. For each stock, count the managers who bought it in a quarter and the ones who sold it. Then ask how lopsided that split is compared with what chance would give, given how many buyers there were across the market that quarter. Whatever is left over is herding.

For US pension funds they found about 2.7%. Russ Wermers later got about 3.4% for mutual funds. Small numbers. Most of what looks like crowding is managers fishing in the same pond.

I ran the same test on the quarterly 13F filings of 47 to 49 famous managers, among them Warren Buffett, Seth Klarman, Howard Marks, Stanley Druckenmiller and Li Lu, over the five quarters from Q2 2025 to Q2 2026. Each manager gets one vote per stock, and I only kept stocks that at least five of them traded. Share classes count once. Sixteen managers hold more than one line of Alphabet, and my first pass counted them twice, which briefly had Alphabet owned by 43 of the 47.

| Quarter | Stocks tested | Herding, all trades | Among buyers | Among sellers |
| --- | --- | --- | --- | --- |
| Q2 2025 | 106 | 0.6% | −2.5% | 3.2% |
| Q3 2025 | 105 | 2.6% | 0.0% | 3.8% |
| Q4 2025 | 121 | 4.9% | 1.3% | 6.7% |
| Q1 2026 | 139 | 2.2% | −1.3% | 5.0% |
| Q2 2026 | 135 | 3.9% | 3.0% | 3.6% |

Overall it averages just under 3%. Famous or not, these managers herd about as much as the pension funds did in 1992, which is to say not much.

### **The first version of that table was wrong**

My first run made the selling side look far more dramatic. Herding among sellers came out as high as 8.7%, and among buyers it was negative in three quarters out of five. It was a very clean story.

Part of it was the method. The usual way to split the measure into buyers and sellers applies one correction for chance to both sides, and when buyers outnumber sellers overall, as they did in every one of these quarters, that correction leans. I simulated a world with no herding at all, same number of managers per stock, and chance alone was worth up to two points on the selling side. The table above corrects each side separately. The gap got smaller. It didn't go away.

### **They agree on what to leave**

After the correction, herding among buyers averages 0.1% across the five quarters. Among sellers it averages 4.5%, and it's positive in every quarter.

Here's what the one-sided trades looked like in Q2 2026: stocks at least five managers traded, where every one of them went the same way.

| All sold | Managers | All bought | Managers |
| --- | --- | --- | --- |
| Oracle | 6 | Danaher | 8 |
| Synopsys | 5 | Carvana | 6 |
| MasTec | 5 | PepsiCo | 5 |
| Coherent | 5 | Sysco | 5 |
| Sandisk | 5 | Performance Food Group | 5 |
|  |  | Workday | 5 |

The buy list is longer, and that's honest: Q2 2026 was the quarter where the two sides came closest. I left out an index fund and SpaceX, which shows up in these filings for the first time this quarter, so nobody could have sold it.

I don't know why exits are more coordinated than entries. My guess is that reasons to leave tend to be shared, a guidance cut or a valuation everyone can see, while reasons to buy are private theses, and there are as many of those as there are managers. It's a guess. Five quarters isn't much data.

### **The biggest names split down the middle**

The most widely held stocks were traded in both directions at once.

| Stock | Held at quarter-end | Added | Cut or sold out |
| --- | --- | --- | --- |
| Alphabet | 26 | 7 | 12 |
| Meta | 21 | 7 | 7 |
| Visa | 20 | 5 | 8 |
| Microsoft | 18 | 8 | 9 |
| Amazon | 17 | 7 | 8 |
| TSMC | 14 | 6 | 7 |

For the ten stocks that at least ten managers traded in Q2 2026, the measure was negative, −2.4%. They disagreed more than coin flips would. A year earlier the same kind of group read +8.1%, though, so ten stocks can't carry a general rule.

### **What it can't tell you**

Forty-seven famous managers is a hand-picked group, chosen because they're famous, and that's a bias of its own. The filings show positions at quarter-end, so anything bought and sold in between is invisible. They also arrive up to 45 days late, so none of this is news by the time anyone reads it, and none of it is a reason to buy or sell anything.

The numbers come from our own database, which is a fair objection. They're also checkable. The [full dataset is public under a CC BY 4.0 licence](https://talval.com/superinvestors/dataset), every manager and every position, and the test is a few dozen lines of Python.

---

Razvan Luca is the founder of [Talval Research](https://talval.com), an independent equity research platform covering around 950 US and European listed companies. Talval publishes valuation estimates and tracks the disclosed portfolios of 49 well-known investors, released as an open dataset.