It is hard to isolate human cooperation features outside the prevalent experimental laboratory context in a controlled setting. Since cooperation rates are very context-sensitive, a new setting-a public goods game in an online game-is analysed. This unique dataset addresses multiple critical issues: it is more realistic, individuals have an intrinsic motivation to play, there are no observer effects, and data are available for more than 18 000 individuals playing over 10 months. Participants come from 10 countries, which allows us to test for cultural influences on cooperation. We find small differences in cooperation rates between countries, ranging from 8.5% (Argentina) to 14.1% (Greece). Moreover, cooperation remains stable over long periods of time. Different cultural, economic and religious backgrounds do not seem to have a discernible influence on cooperation rates. Instead, individual differences seem to play a larger role. Finally, cooperation levels may be lower than suggested by experiments from the classical laboratory context.
Keywords: cultural differences; human cooperation; laboratory context; long-term cooperation; observer effect; public goods game.