The role of correlation in two-asset games : some experimental evidence

In our experimental setting, participants face the decision to invest into two assets which are subject to correlated information. While fundamental states and signals about fundamental states are correlated, success and default of the investment projects is determined separately. Nevertheless, correlation of signals may give rise to spillovers through informational contagion since participants may overvalue correlated signals resulting from a double-counting problem in the updating process or may be prone to behavioral biases related to good and bad news. Quite strikingly, in our setting, the degree of correlation does not promote pronounced contagious effects. In particular, this is consistent with the theoretical two-dimensional global games solution of the underlying investment game. However, a heuristic of neglecting correlation and signals about the second asset has also merits to explain participants' investment behavior. In some treatments we can distinguish between participants' strategies being derived from the twodimensional global game and from a heuristic being derived from a onedimensional game. We cannot reject that people play the two-dimensional investment game as it would be two separate one-dimensional games and ignore correlation.