Most models of strategic information transmission fall in one of two categories. Following Spence (1973) the first category focuses on costly signaling. The second category consists of cheap talk games, with Crawford and Sobel (1982; CS) making the pioneering contribution. In this paper, we investigate which channel will be predominantly used for information transmission in an experimental setting that allows both channels. In three treatments, we vary the dis-alignment between the sender’s and receiver’s preferences. Even though subjects sometimes choose signal costs when preferences are less aligned, information transmission is primarily shaped via cheap talk. Subjects separate with signal costs in an additional treatment where the signal costs provide the only communication channel, but they immediately switch back to using cheap talk once the possibility to use this channel is introduced.
Joint work with Theo Offerman and Randolph Sloof (University of Amsterdam).