Among the more notable studies published in recent years was a paper by Hendrik Bessembinder titled “Do Stocks Outperform Treasury Bills?” His key finding: Between 1926 and 2016, just 4% of all stocks accounted for all of the net gains in the U.S. market. The other 96%, as a group, delivered returns that were no better than Treasury bills, which returned just 2% per year over that period.

It was a surprising result, and the implication was that diversification is even more important than most investors realized, because a portfolio that missed out on just a handful of the market’s best performers—the next Apple or Amazon, for example—could suffer material underperformance. The solution, of course, was easy: Because they cast such a wide net, total-stock market index funds ensure that investors won’t miss out on the market’s next set of star performers.

But Bessembinder’s research posed a problem for investors interested in an investment strategy known as direct indexing that has been growing in popularity. If you’re not familiar with it, this is how direct indexing works: Suppose an investor wanted to own the S&P 500 but preferred to avoid certain stocks or certain industries—tobacco, for example. With a direct indexing service, this investor could purchase all of the individual stocks in the S&P 500 with the exception of the two cigarette makers (Philip Morris and Altria). The portfolio would then consist of the other 498 stocks. This is just one example. With direct indexing, investors can customize portfolios along dozens of dimensions, screening out companies they prefer not to be associated with.

Direct indexing has been around for years, but because of the volume of trades, and the associated brokerage commissions, required to build a portfolio, it was too expensive for all but the wealthiest investors. That changed in 2019, when a number of brokers dropped commissions on stock trades. Since then, a number of firms including Charles Schwab, Fidelity and Vanguard, have all rolled out direct indexing platforms. This created a price war, which has made direct indexing accessible to more individual investors.

Investors interested in direct indexing, however, have faced a bit of a conundrum because of Bessembinder’s finding that just a tiny fraction of stocks are responsible for essentially all of the market’s returns. Investors worried that if a direct indexed portfolio happened to exclude one of the market’s highflying 4%, the performance penalty could be steep. Imagine, for example, if a portfolio had excluded Nvidia over the past five years, when it has returned more than 1,700%. Through that lens, direct indexing looks potentially risky. But how significant is that risk? This has been an open question, but a recently-published paper, “Exclude with Impunity” by Yin Chen and Roni Israelov, provides more clarity.

Chen and Israelov employed a methodology known as backtesting to assess the risk posed by direct indexing. Using stock market data for a nearly 60-year period ending in 2021, the researchers compared the returns of the overall market—as measured by the 1,500 largest stocks—to the hypothetical returns of a portfolio that excluded some number of those stocks. They looked first at the results of excluding just one stock, then five, 10, 30, 50 and more, all the way up to 500. They then repeated the exercise 1,000 times, with a different set of stocks excluded each time.

The results were counterintuitive: Among the 1,000 scenarios tested, the median portfolio delivered results no worse than the overall index. This was true, surprisingly, even when several hundred stocks were excluded. However, and importantly, in “bad luck” scenarios—where, for example, more than one Apple or Amazon was excluded—the results did indeed trail the index. The impact, though, was far more modest than Bessembinder’s work might have led us to believe. With 100 stocks excluded, the worst decile of bad-luck scenarios resulted in a mere 4% reduction in portfolio value over a 58-year period. It was an almost immaterial difference.

There are some important caveats, though. While the impact of excluding 100 stocks was surprisingly modest, the results did deteriorate in the bad-luck scenarios when more stocks were excluded. When 300 stocks were excluded, for example, the ending portfolio value was about 10% lower than if no stocks had been excluded. Chen and Israelov also looked at the impact of excluding entire industries. There, the range of results was wider, because the relative performance of industries is so different.

The bottom line: If you’re considering direct indexing for your portfolio, this new research should provide a degree of comfort. Suppose you’d like to exclude a handful of stocks or perhaps a few industries from your portfolio. The data indicate that you might not incur much of a performance penalty, if any at all. The key, however, is to be sure the exclusions you choose aren’t too numerous. Want to exclude the two tobacco companies in the S&P 500? That’s unlikely to dent performance much. According to the “Exclude with Impunity” data, you could leave out as many as several dozen stocks (from an index of 1,500) without introducing too much risk. But as a rule of thumb, looking at the data, I would draw the line at 100. Or if you go the route of excluding entire industries, I wouldn’t exclude more than four (out of 49 total). This should limit the performance impact, even if one of the stocks excluded turns out to be one of Bessembinder’s star 4%.

If you’re still on the fence about direct indexing, there’s one more factor to consider: Direct indexing typically delivers a tax benefit which may help offset any impact from excluding stocks. Because a direct indexed portfolio consists of individual stocks, it’s much easier to sell tax-efficiently. Imagine, for example, if you purchased shares in an S&P 500 index fund 10 years ago. If you wanted to sell some of those shares today, you’d be somewhat hemmed in. Because of the market’s strong run, the S&P 500 is up about 220% over the past 10 years, meaning you’d incur substantial gains on each share sold.

If, on the other hand, you owned all 500 stocks in the index individually, then you’d have far more flexibility. That’s because of the nature of averages. While the index has gained 220% overall, approximately 250 stocks have gains smaller than that. Indeed, 30 stocks actually have losses over the past 10 years. So if you owned these 500 stocks individually and were looking to sell shares, you’d have much more of an ability to control your tax bill. At the same time, if you were charitably inclined, you’d be able to select the most highly appreciated shares to donate to charity, thus sidestepping any capital gains impact.