There is no commonly accepted definition of a replication. TRN defines a replication as any study whose primary purpose is to establish the correctness of a previous study.

There are many gradations of replications, stretching from pure reproduction of key finding(s) of a previous study; to checking the robustness of those findings to changes in data, estimation procedure, model specification, etc.  There is subjectivity in deciding when a replication becomes an original study in its own right.

Why replications? They enable researchers to (i) determine the fragility or robustness of previous research findings, and (ii) identify why studies reach different conclusions on the same subject.