I could not replicate all results from a paper (replication data were made available online), and asked the author for advice. This is the reply:
Let me check and compare the datasets on the web and the one that I have, and I try to get back to you soon.
One hour later I received this detailed email
I just quickly ran the ordered probit (or logit) regression (with time dummies) and got the same outcomes. I will check dataset more closely to see whether I uploaded the wrong dataset.
By the way, did you use “ordered logit” or just OLS? For all the outcomes (…) we used (…).
For the year dummies, in our dataset (…)
For temporal dependence issue (time correlation), we (…)
Anyway, I will try to compare the datasets and get back to you soon. (…)
More replication correspondence
I collect more responses from original authors here.