Tax evasion is a pervasive problem in many countries. In particular, some developing countries do not collect even half of what they would if taxpayers complied with the written letter of the law. The academic literature has not been oblivious to the need to explain why people pay (or do not pay) taxes. However, the empirical literature has not yet reached consensus. This paper reports the results of a large field experiment that tried to affect compliance by influencing property tax taxpayers' beliefs regarding the levels of enforcement, reciprocity, and peer-effects of the tax system in a municipality in Argentina. Results indicate that those taxpayers that received the deterrence message have a higher probability (almost 5 percentage points) to comply than the taxpayers in the control group. No average effects are found for the other two treatments. However, this average effects masks important results. After receiving the reciprocity and the peer-effects messages, the probability of compliance increased for some taxpayers but it decreased for others according to their underlying distribution of beliefs. The evidence in this paper advances the state of knowledge, may help to reconcile some of the results in the literature, and provides the basis for advancing policies and research on tax compliance in developing countries.