I examine an epidemic model of noise trading, where personal interaction within social groups results in herding behavior and informational cascades, as suggested by Shiller (1995). It is proposed that demographic characteristics such as population density will determine the level of interaction within like-minded social groups. The resultant herding within groups and divergence of opinion between groups will affect the intensity of stock trading. In a study of 49 countries, I find that stock market turnover is higher in countries with denser population and an urban population which is distributed across several cities, after controlling for the quality of legal and political institutions and per capita income. The degree of religious polarization in the country does not, however, appear to affect the stock trading intensity. Also, the effect of population density seems attenuated in economically developed countries. This is consistent with the idea that group ties become weaker in more complex societies and thus the impact of social interaction on individual investor decisions is smaller.