In this paper we aim at investigating whether introducing more transparency may help reducing embezzlement. This is done by running a series of laboratory experiments based on a modified dictator game with intermediaries who have the opportunity to embezzle money between the donor and the final receiver.
Precisely our design is a 2x2 design. We vary both the number of intermediaries (from no intermediary, short chain with one intermediary to a long chain with two intermediaries) as well as the degree of information of the final receiver about the initial amount sent by the sender (no information or full information).
We ran our experiment in Tanzania, considered as one of the most corrupted countries in the world according to Transparency International. Our subject pool consists of 1080 participants from two locations in Tanzania : Dar el Salaam and Mazimbu.
Our main findings indicate that reducing the number of intermediaries increases significantly the average amount received by the final receiver.
This is due to the combination of two effects i) the fact that there are mechanically less potential embezzlers and ii) the fact that the senders are significantly more generous when the number of intermediaries is low. Regarding the effect of information, this effect is less clear cut. W find no evidence of an information effect on the amount received by the final receiver. While information disciplines the intermediaries, this is only the case when the chain is short (i.e. with one intermediary).