Just half of U.S. households invest in stocks in some form, whereas traditional financial economic models suggest near-universal participation. This discrepancy describes the “participation puzzle,” which the authors explore by investigating the intelligence quotient (IQ) as a primary driver.

Two large provinces around Helsinki form the basis of an expansive dataset. IQ data are obtained from nearly 160,000 Finnish Armed Forces (FAF) intelligence assessment tests taken between 1982 and 2000 by males at commencement of their compulsory national service. IQs are categorized on the stanine scale of one to nine. The Finnish Tax Administration provides income, wealth, demographic, and address data from year 2000 tax returns. Household trade data are obtained from the Helsinki Stock Exchange (HEX) for 1995 to 2002, with year 2001 returns used to compute portfolio variance.

Summary statistics show that average stock market participants have higher IQs (by one stanine), earn 50 percent more income, are more educated, and are more likely to be married and have children than nonparticipants. Additionally, income, wealth, and the ratio of those with the highest to those with the lowest educational attainment increase monotonically with IQ. Participation increases from 10 percent to 47 percent between the lowest and highest IQ extremes.

The authors study the marginal effect of IQ on participation by introducing 67 control variables and performing probit regressions. Participation increases monotonically with IQ, and they observe a statistically significant difference of 20 percentage points between the extreme stanines. IQ differences represent the greatest explanatory variable, which are 20–30 percent larger than those explained by income differences. All three subcomponents of IQ influence participation, with a t-statistic of at least 32. The results are robust to the measurement year, when the authors omit blocks of control regressions, age-filtered samples, Nokia holdings, and employer stock holdings.

To exclude any possibility of participation costs driving the IQ–participation relationship, the authors next restrict the test to include only the most affluent. Within the top decile of income/wealth, there is a 16–23 percentage point participation rate difference between the extreme IQ stanines, which is comparable with the full sample findings.

To assess secondary effects, such as IQ leading to wealth, income, and so on, the authors use the Blinder, Oaxaca, and Fairlie method. This method shows that secondary effects account for around 64 percent of the 37 percentage point difference in participation between the highest and lowest IQ stanines, with wealth, education, and income being the main conduits.

The FAF data cover males only. Address information, however, allows sisters to be identified. Using brothers’ IQ as a proxy, the analysis is repeated on more than 4,000 sisters. The authors observe a similar, if less significant, IQ–participation relationship. The robustness of the technique is verified by identifying brothers whose IQs are already known. They also find that endogeneity, which is the possibility of shared family knowledge, is important when using brother pairs.

After establishing that IQ level is a strong driver of stock market participation, the authors investigate possible causes. By inference from the HEX data, they find that low-IQ investors earn statistically significantly lower Sharpe ratios, mainly driven by higher diversifiable risk levels. A 3.5 percent increase in return variance per one-stanine drop in IQ is observed. Additionally, high-IQ investors own more stocks and have more exposure to high book-to-market and small stocks, which are well established as returning higher Sharpe ratios. For a 30 percent risk portfolio, bottom-stanine IQ investors earn at least 21–33 bps less annually than top-stanine investors, justifying the supposition that low-IQ investors rationally opt out of stock market investing.

The authors conclude that IQ is an important determinant of stock market participation—and rationally so, given likely return differences. They proffer that the compounding effect of higher returns may be as important in explaining wealth differences across IQ levels as is earned income. Finally, the authors hypothesize that markets may be more efficient than previously assumed because participants are smarter than average.