The ability to cooperate is a central condition for human prosperity, yet a trend of declining cooperation is one of the most robust observations in behavioral economics. The massive replication of declining cooperation has almost exclusively been carried out in student populations, which opens up for the question of whether the declining cooperation is predictive for the population at large. I make two steps to address this knowledge gap about cooperation stability in the general population. First, I measure repeated cooperation among students and a representative sample. Among the students, I confirm the usual decay effect of cooperation. However, among the non-students, the behavior is hugely different and approaches no decay. Secondly, I stress test the cooperation stability among non-students by manipulating the composition of preferences so that fast decay and no decay are predicted. I observe that the cooperation stability is remarkably unaffected by this manipulation.