When Research Promises the Impossible, Investors Pay the Price

Based on research by Jefferson Duarte, Christopher S. Jones (University of Southern California), Mehdi Khorram (Iowa State), Haitao Mo (University of Kansas), Junbo L. Wang (Louisiana State)

Key findings:

• Many “can’t-miss” option-based trading strategies examined in academic research look impressive on paper because they rely on future data — something no real investor can actually use.
• Once the time-traveling data is removed, those sky-high returns shrink and the risks start to show.
• The pressure to publish big results may help explain why those mistakes were made in the first place.
   

If something seems “too good to be true,” it probably is — and this old axiom is amplified when money is on the line. According to new research by Jefferson Duarte, the Gerald D. Hines Associate Professor of Real Estate Finance at Rice Business, investors should be skeptical of investment strategies that promise big returns with little risk — even when they’re published in academic journals.

We all know the sordid tale of Bernie Madoff. The former New York City asset manager lured investors returns that appeared to far outpace the market. For years, he attracted a steady stream of clients with a seemingly flawless performance record. In reality, investor withdrawals were paid using money from new deposits. When the 2008 financial crisis shook markets, Madoff’s $65 billion Ponzi scheme collapsed. Investors rushed to withdraw funds, but there wasn’t enough cash to go around.

Unfortunately, unrealistic performance claims aren’t limited to bad actors like Madoff. Even some peer-reviewed finance papers describe strategies with returns that appear implausibly high.

For an article forthcoming in The Review of Financial Studies, Duarte and his co-authors dug into the details of several influential empirical options papers. And when they tried to replicate their results, they couldn’t. The authorship team found that the option-trading strategies proposed in these papers exhibited “look-ahead bias” — a statistical error that arises when researchers use information that would not have been available when the trading decisions were actually made. 

“When we eliminated the intraday data from the analysis, the returns were much smaller, and the level of risk also increased. These were not extremely good deals anymore.” 

In layman’s terms, the options trading papers used future data — for example, the average price of a stock or option over the course of a day — to make trading decisions that were supposed to happen at the start of that same day. That’s a problem because, in real life, traders don’t know what the average price will be until the day is over.

So, the authors of these influential empirical options papers were effectively “looking ahead,” using information from the future to make their predictions. This made the results appear far stronger than they would be in reality — essentially giving the strategies an unfair, time-traveling advantage.

One influential study published several years ago claimed that a simple options-based approach could generate astonishing daily mean returns of more than 3%. The strategy involved buying and selling call options — contracts that give investors the right to purchase a stock at a specified price.

To decide which call options to trade, the researchers used filters based on liquidity — how easily a contract can be bought or sold without affecting its price. But their method relied on each contract’s average price over the entire trading day to make selections at the start of that same day — using information that wouldn’t actually be known until after markets closed. When Duarte and his co-authors replicated the strategy using only information available at the start of the trading day, the returns proved far more modest. 

“Just imagine profiting a few percentage points per day,” says Duarte. “If you could do that every day for years, you would have a money wealth machine. But in practice, you would need a crystal ball to do this. When we eliminated the intraday data from the analysis, the returns were much smaller, and the level of risk also increased. These were not extremely good deals anymore.” 

But why was a paper with such a significant error published in the first place? Publication bias within academia could have contributed. Academic journals favor research that has very strong results, and this can nudge authors to find ways to make their results look stronger than they actually are. 

Simply put, if a proposed options-trading strategy did not generate exceptionally high returns, it might never have been published. But a strategy showing smooth gains averaging more than 3% per day is bound to attact attention — even if it isn’t actually feasible. 

Duarte’s research examined only three options papers, and he believes there may be more cases of look-ahead bias in the empirical options literature. One concrete way researchers can improve on this is to provide a clear, detailed description of their methods so others can replicate the results. 

To reproduce the findings in one study, Duarte and his co-authors had to test 190 different procedures just to figure out how the sample had been selected. If published papers were easier to replicate, he says, it would be far simpler to detect look-ahead biases in the first place. 

Duarte also argues that researchers should exercise restraint when selecting their sample. While the use of filters is common in options research and can help remove data errors, they also create opportunities for mistakes — such as look-ahead bias — to slip in.

And if any options strategy appears to outperform the market by a wide margin, it should invite additional scrutiny. Exceptional risk-adjusted returns may signal that an error in the empirical analysis has gone unnoticed.

“When you talk to practitioners — people that actually implement these option strategies,” Duarte says, “they will tell you that these high risk-adjusted returns simply can’t happen on a consistent basis. When they try to replicate a strategy like this, they can’t do it either.”

Written by Ty Burke

Duarte, et al. “Too Good to Be True: Look-ahead Bias in Empirical Options Research.” Forthcoming in The Review of Financial Studies (2025).