We analyze students' cheating behavior during a national evaluation test. We model the mechanisms that trigger cheating interactions between students and show that, when monitoring is not sufficiently accurate, a social multiplier may magnify the effects on students' achievements. We exploit a randomized experiment, which envisaged the presence of an external inspector in the administration and marking of the tests, to estimate a structural (endogenous) social multiplier in students' cheating. The empirical strategy exploits the Excess-Variance approach (Graham, 2008). We find a strong amplifying role played by social interactions within classrooms: students' cheating behaviors more than double the class average test scores results. The effects are found to be larger when students are more homogeneous in terms of parental background characteristics and social ties.