Using a laboratory experiment, we investigate the effect of gender composition on the group dynamics of microfinance clients.

In a novel face-to-face problem solving game, we find that female-only groups outperform both male-only and mixed groups.

Similarly, when making joint risk-decisions in groups, females-only groups take more risks than male.

However, in the public-goods game we find no robust evidence of female groups contributing more than male and mixed groups.