At the heart of financial research lies a seemingly simple question: what is the risk-return profile of stocks and corporate bonds? New research based on the Belgian stock market from 1850 to 1913 shows that the answer changes fundamentally when illiquidity is taken into account.
The figure below illustrates the difference between traditionally observed returns and returns corrected for selection bias. That is, it accounts for the fact that in historical markets, many assets simply were not traded. Without a transaction, there is no price—and therefore no return—to record. These “invisible” assets—often companies in distress or bonds with no buyers—are structurally missing from our data. The returns we do observe are inevitably biased: they reflect only the “successes,” not the full universe.
Figure 1: Stocks
The results are striking. Corrected returns are systematically lower than observed returns. For stocks, this means the average annual return drops from 4.97 percent to 2.66 percent. For bonds, the impact is even more dramatic: from 1.77 percent to a mere 0.28 percent. This implies that much of the presumed outperformance of certain asset classes—such as corporate bonds—disappears entirely once adjustments for illiquidity are made. In other words, the traditional view of bonds as a relatively safe and profitable alternative to stocks in the nineteenth century is partly an illusion.
The classic risk-return principle is also called into question. Modern financial theory holds that investors are only willing to bear risk if they are compensated with higher expected returns. In practice, this is usually measured via beta, a metric for systematic risk. The study behind the figure below shows that this relationship holds only when using uncorrected data. Once returns are adjusted for selection effects, the positive link between beta and average return vanishes. High-risk assets no longer yield more than low-risk alternatives. In fact, the difference in Sharpe ratios—the ratio of return to risk—drops substantially and is almost completely neutralized for bonds.
Figure 2: Corporate bonds
This has direct implications for institutional investors. Many allocation models, historical scenario analyses, and long-term forecasts are based on datasets that reflect only the transactions that actually took place. But just like in the nineteenth century, what is traded today is only a selection of what exists. Especially in segments such as private equity, infrastructure, or emerging markets, illiquidity and selection bias are pervasive. Investors who fail to account for this overestimate their expected returns—and underestimate their actual risks.
The lesson from the figures above is simple, yet profound: return is not a neutral observation, but a consequence of what can be seen. And what is seen depends on who is trading. In illiquid markets—historical or contemporary—this can lead to structurally flawed assessments of risk, performance, and value. By correcting for what is missing, we discover that the return we thought we knew may never have existed at all.
Gertjan Verdickt is Assistant Professor of Finance at the University of Auckland and a columnist at Investment Officer.