The intuitive link between legal insider trades and stock returns proves elusive as a quant strategy.

When most people hear “insider trading,” they think about the illegal kind: someone in a business suit and handcuffs being led ignominiously from a Wall Street office, having bought or sold shares after becoming privy to some non-public information they heard.

But trades can be made legally by insiders. After all, many executives and employees receive part of their compensation as shares in their company, and they need to be able to sell them at some point. In connection with such insider trades, the transaction generally has to be reported to the securities regulator; in the U.S., that means filing the report with the Securities and Exchange Commission (SEC) within a short period of time following the transaction. In addition, most companies adopt strict policies on when such transactions can occur, such as moratoria on trading around public earnings announcements.

Understandably, the existence of the intelligence around insider activity has long intrigued investors and finance researchers. One might expect these buyers and sellers to have better knowledge of a company’s immediate prospects than the market at large, and the fact that they have to report their transactions potentially gives an investor insight into what they are thinking about their companies at any moment in time.

If insiders are buying, surely it makes sense to buy with them? And if they are selling, shouldn’t we be selling, too?


Over the years, a number of research papers have claimed to show a link between insider activity and subsequent risk-adjusted outperformance of the shares concerned. Most acknowledge that the link is weak when the insider-trading data is used bluntly or naively, and make their case by adding extra layers or filters to that data.

For example, the authors of one 2012 paper1 looked at companies where there were intensive insider purchases and recent announcements of share buyback programs, and found that the link between these factors and subsequent stock performance is evident only for high-value stocks. A paper2 from 2011 found that the trades of Chief Financial Officers tend to perform better than those of Chief Executive Officers, perhaps because CFOs know more about the firm’s finances, perhaps because CEOs are under more regulatory scrutiny—or naturally overly optimistic! Six years earlier, another paper3 claimed that insider trades had predicted future returns, but also that insider purchases predicted improved returns even when equity analysts had downgraded their expectations for the company’s shares. We find these papers to be indicative of what we see in academic research into this topic.


We are often able to reproduce the results of these academic exercises, including when we test the hypotheses using market performance data up to the present day—despite evidence that the predictive power of naively constructed insider-trading signals appears to have grown much weaker over the past decade. 

But an academic exercise is not the same as practical implementation of a quantitative investment strategy. Two problems keep recurring:

First, the extra filters that academic researchers apply substantially reduce the number of stocks in play. On average, value companies undertaking a share buyback program and showing intensive insider purchasing account for less than 1% of U.S. stocks, for example. CFOs make up only 7% of insider trades—and we have found that adding other top insiders into the scoring muddies the signal.  

Second, the link between trades by insiders and stock outperformance has decayed through time. Perhaps this is due to improving market efficiency and widespread access to the publicly available insider transaction data, which illustrates the need for quantitative managers to constantly adapt. More pessimistically, it may be an indictment of the academic research process, which has incentives to mine signals from historic data that may not work as effectively out-of-sample.

“Insider Boost”

One way to illustrate both of these effects is to apply the insights of the academic literature with a trading strategy that incorporates what we call an “Insider Boost” metric that changes a stock’s place in a quantitative, multifactor ranking.

A stock receives a binary “Insider Boost” if there have been purchases totaling $100,000 or more in the past 90 days by top insiders (excluding CEOs), and the stock is identified as high-value by standard value metrics. That boost moves the equity up 30 percentage points in the rankings: for example, in a universe of 1,500 stocks, an equity ranked number 500 (ranked 33rd percentile) would move up to 50 (ranked 3rd percentile). The chart below shows the marginal information coefficient attributable to the “Insider Boost”—in other words, the difference between the information coefficient of a standard value score with and without the boost being applied.  

Marginal information coefficient from using the “Insider Boost”, 2006 – 2017

Source: Bloomberg, S&P Capital IQ, Neuberger Berman. The chart shows monthly marginal information coefficient contribution by the “Insider Boost” metric, which increases a stock’s ranking number by 30% when there have been purchases totaling $100,000 or more in the past 90 days by top insiders (excluding CEOs), and the stock is identified as high-value by our standard value metrics. The universe of stocks is the top 1,500, by market-capitalization, of the Russell 3000 Index. Trades are made seven days after the transaction date, taking the closing price of that day.

The results highlight the issues above. Over the first half of the dataset, our “Insider Boost” improves the information coefficient by just 0.07%, on average. In part this is because of the limited number of stocks that get an “Insider Boost.” Since 2012, it has made no discernible contribution at all. Over the entire period, the marginal value of adding an “Insider Boost” is positive, but negligible.

Perhaps counterintuitively, insider activity appears to give us little actionable insight into the future performance of the stocks involved. This may be because insiders are often banned from transacting during periods when we might expect stocks to be re-rated by the market. It may be because a lot of the transacting is not profit-maximizing. And certainly this is an example of the common effect whereby a well-known signal deteriorates through time. Maybe further academic research would show stronger predictive results from illegal insider trading—but handcuffs are never a good look.