One of the errors the replication found was that the STATA code of the original study had matrix CLOSE_D = J([_N], 12, 1000) when it should have been matrix

One of the errors the replication found was that the STATA code of the original study had matrix CLOSE_D = J([_N], 12, 1000) when it should have been matrix CLOSE_D = J([_N], 75, 1000) When our public policy is based ultimately on results produced by code, I think we must train our social scientists more systematically in programming ("no magic numbers") and have more public scrutiny of the code (this replication happened 11 years after the original study).