Fig 1.
Predicted effects for the interaction of valence, source, and reliability in comparison to the baseline with no third-party information.
Fig 1 shows the predicted relative effects on trust for third-party information under different combinations of source and reliability. The triangles on the left side indicate increasing effect sizes in terms of placing trust (+) and withholding trust (-). Each line represents third-party information with a specific set of characteristics in terms of the source and reliability of the information. When both source and reliability operate in the same direction, effects should be relatively large in magnitude (e.g. friend tells experience) or relatively small in magnitude (e.g. stranger tells gossip). When source and reliability work in opposite directions (e.g. friend tells gossip or a stranger conveys an experience), relative effects cannot be predicted a priori. The baseline involves no third-party information.
Table 1.
Overview of the complete 2x2x2 vignette set and the baseline vignette with no third-party information.
Fig 2.
Example vignette of a friend of the buyer (participant) sharing a negative experience with the seller.
Fig 3.
Reference category is the vignette ‘No third-party information’.
Table 2.
Valence of information.
Table 3.
Reliability of information.
Table 4.
Source of information.
Fig 4.
The average marginal effects on the trust decision.
In Fig 4 the vertical axis shows the predicted relative effects on trust of the valence (negative versus positive), source (stranger versus friend), and reliability (gossip versus experience) of third-party information. All eight possible combinations are shown, and they are ranked according to the predictions in Fig 1. As discussed in the caption to Fig 1, the cases in which source and reliability operate in opposite directions have no clear ranking, and here we have simply chosen to posit a stronger effect for a friend with gossip compared to a stranger with experience. The horizontal axis shows the average marginal effects (AMEs) with confidence intervals as estimated from our data. The vertical line at zero represents the baseline category in which no third-party information was available.