Help - Moderation with dummies in Cox Regression


I am currently working on my masters thesis.

I am testing how team change following an acquisition affects the survival of the acquisition itself (expecting a negative effect) and to do so I am using a Cox survival analysis. I have 2 moderators which a are 1/0 dummies, one is internationality (I expect that if the acquisition is international the effect of team change will be even worse) and inexperience (again, if the company has never been in a certain market, the effect is going to be worse).

To do the moderation I multiplied team change (centered) values with the dummies and I get the results of table 1 (in the attachment). Now these results are completely counterituitive: if team change increases the hazard and so does internationality (even if insignificant), how is it possible that when combined they appear to have a very positive effect on survival? This does not make sense to me.

I experimented changing the signs in the moerator and suddenly the results are much more consistent with each other (table 2). Now, I did this just following my instinct (I am really not a stats genius), but does it make sense? It does for my research, but I am not sure it does for statistics.

I would really appreciate any help!