PTDo investors have an accent?
Trusting yourself too much, holding on to the stock that keeps falling, following the herd: investors all over the world make the same mistakes. With daily data from 64 stock exchanges, from 2010 to 2024, the thesis shows that each country's culture changes the size of these biases, and that the pandemic even reversed the direction of some of them. Explained in 12 screens.
Of the 64 exchanges, how many showed each behavior in each period. During the pandemic, "anti-herding" nearly disappeared. Our count from Table 17 (p < 0.10).
Three mistakes almost every investor has made
Classical theory assumes a cold investor who uses all available information. Behavioral finance shows otherwise: emotions and mental shortcuts lead to predictable mistakes, called biases. The thesis studies three of them.
Are these mistakes the same size everywhere? The thesis bets they are not: each country has a national culture, a shared way of thinking and feeling passed down through generations, which can amplify or curb each bias.
The thesis's research question, section 1.1.
Do you recognize the scene? Tap to see the name of the bias and how the thesis measured it:
Six scores from 0 to 100 to describe a country's culture
The thesis uses Hofstede's cultural dimensions, the most widely used model for comparing countries. Each country gets a score from 0 to 100 on six dimensions. Pick one and see where the 64 countries in the study fall:
Each dot is a country; tap a dot to see its name. Countries without a score on the dimension are left out (Jamaica on long-term orientation and indulgence; Israel, Sri Lanka, Tunisia and Jamaica on indulgence). Source: Hofstede Insights (2024), Appendix I.
A young field that has doubled in size since 2017
Before going to the data, the author combed Web of Science and Scopus for everything published on national culture and investor behavior in well-ranked finance journals. 60 articles remained, from 2009 to 2024. Output more than doubled from 2017 onward; the most influential study is Schmeling (2009), and the most studied countries are the United States and China.
Circle area proportional to the number of documents. Filters: English only, only journals ranked by the ABS (level 2 or higher), no duplicates; then abstract screening. Search in May 2024. Source: section 3.3.
What do these articles talk about? An analysis of the abstracts' vocabulary separated three major themes. Tap a theme:
% of abstract segments classified in each theme (201 of 219 segments, Reinert's method in IRaMuTeQ). Questions from Table 5, which guided the later studies. Source: sections 3.5 and 3.7.
Culture has roots in the economy, faith and law
The author compared countries' cultural profiles by development, majority religion and legal origin. The arrows show what changes relative to the reference group. Tap a layer:
Only statistically significant differences are shown (*** p < 0.01; ** p < 0.05; * p < 0.10), from multinomial logistic regressions with 60, 79 and 98 countries with complete data. Source: Tables 8, 9 and 10.
How to measure a bias without asking anyone anything
Instead of questionnaires, the thesis reads biases in the footprints they leave in the market: trading volume and price movements, day by day, sector by sector.
Each market's days become three levels (low, medium or high bias, by quartiles), related to the six dimensions by multinomial logistic regression; a second model (GAM) checks for curved relationships. Data: Capital IQ. Source: sections 5.2, 6.3 and 7.3.
Made-up sectors and returns, only to show the logic. The thesis's measure follows Chang, Cheng and Khorana (2000): if, on days of big moves, the average distance of sectors from the market grows less than expected, there is herding.
During the pandemic, three cultural traits switched sides
Before COVID-19, more individualistic, more competitive (masculinity) and more uncertainty-avoiding countries had more days of overconfidence. During the pandemic, all three began to curb the bias, and stayed that way afterward. For the author, facing a risk nobody controls, people began to act more as a group.
View as table
Height = size of the coefficient (full model, column 7); color = direction. Coefficients per point on the 0 to 100 scale: small one by one, they add up to large differences between countries far apart on the scale. Tap a block. Source: Table 12.
Rich, long-term-minded countries hold on to losses less
The higher the income per person, the lower the chance of days with a high disposition effect. More individualistic and long-term countries also show less of the bias. Contrary to earlier studies, masculinity and indulgence are also linked to less bias.
Full model (column 8), per point from 0 to 100; income, in logarithm, is shown apart (−0.0455***). In the GAM, individualism, uncertainty avoidance and power distance are linked to more bias over part of the scale. Source: Table 15 and section 6.5.
And country by country? Daily average of the disposition effect index (1 = people sell what rises as much as what falls):
Black line: sample average (1.77). Brazil (1.13) ranks 59th of 64, among the countries that hold on to losses least. Daily averages are pulled up by extreme days; medians are close to 1 in almost every country. Source: Tables 14 and H-1.
In the pandemic, markets stopped going against the flow
Before COVID-19, 26 exchanges showed the opposite of herding: investors each acting on their own. During the pandemic, only 9 stayed that way; most became neutral and herding appeared in 14. Afterward, anti-herding returned (25) and herding fell to 7, almost all in emerging markets such as Morocco, Tunisia and Vietnam.
Herding = negative and significant γ₃ coefficient (p < 0.10) in each exchange's regression, as in the thesis's method (section 7.3); anti-herding = positive and significant. Before: Jan 2015 to Mar 10, 2020; pandemic: Mar 11, 2020 to Dec 2021; after: Jan 2022 to Jun 2024. Tap a country. Our count from Table 17.
After the crisis, hierarchy weighed much more
Income per person protects against herding in all three periods. Among cultural traits, power distance (accepting hierarchy) always pushed toward herding, and its weight nearly quintupled after the pandemic. Long-term orientation always curbed the bias. Masculinity and uncertainty avoidance changed sign. Tap a line:
Coefficients from the full model (column 8), per point from 0 to 100; income enters in logarithm, so it sits in the separate panel. The set explains more after the crisis (pseudo R² from 0.128 to 0.139 and 0.236). Source: Table 18.
Pick a country and see its profile
For each of the 64 exchanges, we brought together the cultural scores, the average disposition effect and what the herding test showed in each period. Compare with another country:
Panel put together by us from Appendix I, Table H-1 and Table 17. The thesis's models describe tendencies across countries; they do not predict the behavior of a country or of an investor.
Educate, regulate and invest with culture on the radar
Financial education
Programs and risk warnings work better when adapted to the cultural profile of investors, instead of a single recipe for everyone.
Sections 6.5 and 8Regulators
Special attention to low- and middle-income markets, where rising income reduces herding most, and to hierarchical countries as crises end. Crisis communication that reduces ambiguity.
Section 7.5Companies
Consider the target market's culture when negotiating, entering a country and launching products: where uncertainty avoidance is high, detailed contracts and products that convey safety.
Sections 4.5 and 8Investors
Knowing that the cultural setting weighs on decisions, and that in a crisis this weight can change direction, helps you be wary of your own impulses. Our reading.
Based on chapter 8What this research does not say
- It does not prove cause. It shows associations between culture, income and biases; other country factors may weigh on both. · according to the author
- One score per country. Hofstede assumes a homogeneous culture and misses regions, subcultures and immigrants. · according to the author
- Biases measured by market footprints, not by tests with people; liquidity and trading robots enter the volume. · according to the author
- A single crisis. The pandemic was a health shock; a financial crisis or a war may have another effect, and the period dates are a choice. · according to the author
- Modest explanatory power. The models have pseudo R² from 0.02 to 0.24, depending on the bias: culture is one piece, not the whole explanation. · our note
Credits
IKEDA, Wilson Eduardo. Cultural dimensions and behavioral finance: cross-national evidence of investor biases. Doctoral thesis (Business Administration) — Mackenzie Presbyterian University, São Paulo, 2026. Advisor: Prof. Dr. Denis Forte. Co-advisor: Ana Paula Matias Gama, Ph.D.
Thesis written in English, as five articles, as a dual degree with the doctoral program in Management of the University of Beira Interior (Covilhã, Portugal), with support from CAPES (Brazil's federal agency for graduate education) and FCT (Portugal's science foundation). Data from Capital IQ, Hofstede Insights and the World Bank. This version simplifies the work for a general audience.