I have a problem with data retrieved from database of IMF International Financial Statistics. There are some series that contain observations with note "Identifies the point at which multiple time series versions are linked by butt-splicing". Can I use the whole continuous series with these observations in there for my econometric estimations? If not, how to deal with this data break?
If it's a dependent variable, I would be concerned about crossing that point. What they are saying is that it's one time series up to point T, and another from T+1 on with no attempt to overlap them. It's intended that the two are measuring roughly the same thing or they wouldn't make a continuous series out of them, but the dynamics are likely to be making a rather abrupt transition at the join point.